As a knowledge-intensive and challenging task, automatic generation of long-form wiki-style articles has garnered increasing attention from researchers due to its ability to efficiently integrate, organize and present vast amounts of both structured and unstructured knowledge. To the best of our knowledge, most of the existing mainstream state-of-the-art methods for automatic wiki-style article generation typically follow a "one-shot generation" paradigm: given a topic, (1) first generating a structured outline, (2) then independently and in parallel generating the content of each outline chapter in a one-shot using the chapter title and references. However, the core limitation of the paradigm lies in its disregards inter-chapter correlation and lacks post-generation revision and refinement, resulting in content redundancy, weak relevance and logical inconsistency. To address these issues, we propose WikiREVIEW, a novel multi-perspective review framework for automatic wiki-style article generation. Specifically, our proposed method introduces multi-perspective experts to review the content of each outline chapter at both chapter and paragraph levels following the initial generation, offering evaluation feedback and continuously refining the numerous deficiencies in the initial long-form article, ultimately achieving high-quality wiki-style article generation. Extensive experimental results on the public English dataset FreshWiki and our own constructed high-quality Chinese dataset ChineseWiki, demonstrate that our proposed WikiREVIEW significantly outperforms existing state-of-the-art automatic wiki-style article generation methods across all automatic evaluation metrics and human evaluation.

It is crucial to understand a specific domain by events.Extensive event extraction research has been conducted in many domains such as news, finance, and biology.However, event extraction in scientific domain is still insufficiently supported by comprehensive datasets and tailored methods.Compared with other domains, scientific domain has two characteristics: (1) denser nuggets and events, and (2) more complex information forms.To solve the above problem, considering these two characteristics, we first construct SciEvents, a large-scale multi-event document-level dataset with a schema tailored for scientific domain.It consists of 2,508 documents and 24,381 events under multi-stage manual annotation and quality control.Then, we propose EXCEEDS, an end-to-end scientific event extraction framework by encoding dense nuggets into a grid matrix and simplifying complex event extraction as a nugget-based grid modeling task.Experiments on SciEvents demonstrate state-of-the-art performances of EXCEEDS.Both the SciEvents dataset and the EXCEEDS framework are released publicly to facilitate future research.

Chain-of-Thought (CoT) reasoning is crucial for the performance of Large Reasoning Models (LRMs) but is often hindered by redundant and distracting segments, which incur excessive inference costs and degrade robustness.Existing approaches try to solve this problem by enforcing brevity through external supervision, such as length-based penalties or heuristic truncation.However, these approaches often degrade performance because they disregard the model's intrinsic reasoning dependency and thus fail to distinguish between essential and redundant CoT segments.To address this problem, we propose SGP-CoT, a novel Self-Guided Pruning framework that leverages the model's intrinsic likelihood landscape to identify segments that are extraneous to its specific reasoning pattern.Specifically, SGP-CoT treats the reasoning trajectory as a sequence of semantic units and assesses the necessity of each one via internal likelihood signals, measuring its contribution to the answer and local coherence.Based on this, it selectively removes non-essential segments and then forms high-quality pruning-based preference pairs, enabling the model to learn focused reasoning via self-optimization.Extensive experiments across diverse benchmarks demonstrate that the proposed SGP-CoT significantly reduces output length while maintaining or improving accuracy.These results validate that LRMs intrinsically possess the capability to discern reasoning utility, positioning SGP-CoT as a robust pathway toward scalable inference.

Yanghao Zhou, Haitian Li, Rexar Lin, Heyan Huang, Jinxing Zhou, Changsen Yuan, Tian Lan, Ziqin Zhou, Yudong Li, Jiajun Xu, Jingyun Liao, YiMing Cheng, Xuefeng Chen, Xian-Ling Mao, Yousheng Feng. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

User satisfaction serves as a key indicator of search experience in Generative Information Retrieval (GenIR) systems, and its accurate estimation is essential for system optimization. Extensive research in traditional IR has established interaction signals (e.g., dwell time, clicks, query reformulations) as reliable indicators of user satisfaction. However, prevailing approaches for satisfaction estimation in GenIR (e.g., LLM-as-a-judge) primarily rely on textual content and fail to account for user interactions during the search process. In this work, empirical analysis of real-world data shows that user satisfaction correlates negatively with interaction signals reflecting interaction cost, and positively with response quality. Building on these findings, we propose the Gain-aware Cost-sensitive Satisfaction estimator (G-CoS), an interpretable gain-cost framework for user satisfaction estimation in GenIR. G-CoS models user satisfaction as a dynamic trade-off between Response Quality and multidimensional Interaction Cost. Experimental results demonstrate that G-CoS outperforms LLM-as-a-judge methods, interaction sequence models, as well as machine learning models using the same gain and cost features. Moreover, the learned parameters reveal interpretable associations between gain-cost dynamics and user satisfaction. This work contributes an interpretable framework for user satisfaction estimation and offers insights for GenIR system optimization.

Multi-modal hashing methods have gained popularity due to their fast speed and low storage requirements. Among them, the supervised methods demonstrate better performance by utilizing labels as supervisory signals compared with unsupervised methods. Currently, for almost all supervised multi-modal hashing methods, there is a hidden assumption that training sets have no noisy labels. However, labels are often annotated incorrectly due to manual labeling in real-world scenarios, which will greatly harm the retrieval performance. To address this issue, we first discover a significant distribution consistency pattern through experiments, i.e., the 1-0 distribution of the presence or absence of each category in the label is consistent with the high-low distribution of similarity scores of the hash codes relative to category centers. Then, inspired by this pattern, we propose a novel Distribution-Consistency-Guided Multi-modal Hashing (DCGMH), which aims to filter and reconstruct noisy labels to enhance retrieval performance. Specifically, the proposed method first randomly initializes several category centers, each representing the region's centroid of its respective category, which are used to compute the high-low distribution of similarity scores; Noisy and clean labels are then separately filtered out via the discovered distribution consistency pattern to mitigate the impact of noisy labels; Subsequently, a correction strategy, which is indirectly designed via the distribution consistency pattern, is applied to the filtered noisy labels, correcting high-confidence ones while treating low-confidence ones as unlabeled for unsupervised learning, thereby further enhancing the model’s performance. Extensive experiments on three widely used datasets demonstrate the superiority of the proposed method compared to state-of-the-art baselines in multi-modal retrieval tasks.

Automatic evaluation for Open Domain Event Detection (ODED) is a highly challenging task, because ODED is characterized by a vast diversity of un-constrained output labels from various domains. Nearly all existing evaluation methods for ODED usually first construct evaluation benchmarks with limited labels and domain coverage, and then evaluate ODED methods using metrics based on token-level label matching rules. However, this kind of evaluation framework faces two issues: (1) The limited evaluation benchmarks lack representatives of the real world, making it difficult to accurately reflect the performance of various ODED methods in real-world scenarios; (2) Evaluation metrics based on token-level matching rules fail to capture semantic similarity between predictions and golden labels. To address these two problems above, we propose a scalable and reliable Semantic-level Evaluation framework for Open domain Event detection (SEOE) by constructing a more representative evaluation benchmark and introducing a semantic evaluation metric. Specifically, our proposed framework first constructs a scalable evaluation benchmark that currently includes 564 event types covering 7 major domains, with a cost-effective supplementary annotation strategy to ensure the benchmark’s representativeness. The strategy also allows for the supplement of new event types and domains in the future. Then, the proposed SEOE leverages large language models (LLMs) as automatic evaluation agents to compute a semantic F1-score, incorporating fine-grained definitions of semantically similar labels to enhance the reliability of the evaluation. Extensive experiments validate the representatives of the benchmark and the reliability of the semantic evaluation metric. Existing ODED methods are thoroughly evaluated, and the error patterns of predictions are analyzed, revealing several insightful findings.

Driven by the remarkable progress in diffusion models, text-to-image generation has achieved substantial advancements, underscoring the urgent need for robust automatic quality assessment.This task is inherently complex, requiring evaluations that range from object presence and attribute correctness to relational consistency and visual fidelity.Consequently, current state-of-the-art MLLM-based approaches often rely on powerful commercial models such as GPT-4o, which offer superior reasoning and instruction-following capabilities but are not universally accessible.In contrast, while opensource MLLMs demonstrate promising skills in vision and language understanding, they underperform in comprehensive image quality assessment.To address these challenges, we propose a task decomposition evaluation framework based on GPT-4o to automatically construct a specialized training dataset, breaking down the multifaceted evaluation process into simpler sub-tasks and thus reducing learning complexity.Building on this dataset, we design novel training strategies to distill GPT-4o's evaluation capabilities into a 7B open-source MLLM, MiniCPM-V-2.6,enabling it to better follow instructions across diverse assessment criteria.Furthermore, to reliably and comprehensively assess prior works and our proposed model, we manually annotate a meta-evaluation benchmark that includes chain-of-thought explanations alongside quality scores for generated images.Experimental results demonstrate that our distilled open-source MLLM significantly outperforms the current state-of-the-art GPT-4o-base baseline, VIEScore, with over 4.6% improvement in Spearman and Kendall correlations with human judgments.

3D Visual Question Answering (3D-VQA), which focuses on answering user questions based on a given 3D scene, has attracted increasing attention from researchers. As far as we know, most existing methods first encode 3D point clouds and text separately, then align their features through a fusion layer, and ultimately predict answers based on the fused representations. However, the cross-modal alignment during the fusion stage remains suboptimal, primarily due to the sparsity of 3D point clouds and the lack of 3D scene-text pairs. To address this issue, we propose a 2D-enhanced multi-modal fusion framework for 3D-VQA, leveraging 2D information to supplement the contextual and detailed information missing from the point clouds, thereby enhancing the model's cross-modal alignment capabilities. Specifically, the proposed model consists of three stages. First, it employs a 2D visual-language model to retrieve scene images relevent to the question as supplementary information. Next, a Dual-branch Fusion Module is designed to facilitate bidirectional interaction between the 2D and 3D modalities, enhancing the integration and alignment of visual and textual features. Finally, the fused features are decoded to generate answers. Experimental results demonstrate that our method outperforms existing state-of-the-art baselines.

High-quality datasets are essential for building effective task-oriented dialogue (TOD) systems.The existing TOD datasets often present overly simplified interactions, where users incrementally express straightforward requests that can be managed with basic slot-value style dialogue states, such as "hotel-area = east."However, this approach does not reflect reallife scenarios in which users may express complex constraints and preferences.To address this gap, in this paper, we propose SQLWOZ, a novel TOD dataset designed to capture complex, real-world user requirements.The user requirements in SQLWOZ include the four categories: 1) multiple values for a slot, 2) excluded values within a slot, 3) preferred or prioritized values, and 4) conditional values based on other conditions.We utilize SQL statements as a formalized and expressive representation of dialogue states within SQLWOZ.To evaluate the dataset, we adapt large language models as dialogue agents and conduct extensive experiments on the SQL-based dialogue state tracking, dialogue response generation, and end-to-end TOD tasks.The experimental results demonstrate the complexity and quality of SQLWOZ, establishing it as a new benchmark for advancing TOD research.

Hybrid-modal table understanding (HMTU), which targets leveraging multi-modal table evidence for multi-hop reasoning, has garnered widespread attention. Existing models primarily focus on effectively integrating multi-modal table evidence to enhance the table understanding capabilities of multi-modal large language models (MLLMs). However, these models ignore the fact that different types of table understanding questions lean toward different modalities of table evidence. Consequently, these models suffer from low utilization efficiency and poor interpretability. To address these issues, in this paper, we propose a modality preference alignment model, called ESTJ, which Enhances Structured Tendency Judgment in HMTU. Specifically, ESTJ first samples modality preference data from the responses generated by MLLMs. Then, it alleviates modality preference imbalance by adhering to the principle of least modality priority. Finally, ESTJ performs direct preference optimization (DPO) training based on structured tendency judgment to align modality preference effectively. Experimental results on TableQA and TableFV tasks demonstrate that our proposed model outperforms state-of-the-art baselines. Additionally, these results present fascinating phenomena and unveil profound insights into modality preference for table understanding.

In real-world scenarios, the data usually appears in a streaming fashion. To achieve remarkable retrieval performance in such scenarios, online multi-modal hashing has drawn great research attention due to its high retrieval speed and low storage cost. However, existing online multi-modal hashing methods still fail to achieve satisfactory retrieval performance in the scenarios where the new streaming datapoints all belong to the new classes. Therefore, to further improve the retrieval performance in these scenarios, we propose a novel Prospective Layout-Guided Multi-modal Online Hashing, termed PLG-MOH. Specifically, PLG-MOH first establishes the layout of the Hamming space by generating a series of hashing centers to split the space. Each hashing center will be gradually assigned to a new appearing class, and these assigned centers correspond one-to-one with the classes. Moreover, we propose a novel prospective layout-guided loss, which leverages all the hashing centers, including those not yet assigned to the classes, to supervise the training of hashing model. As the unassigned hashing centers will be designated to the new classes emerging in the future, it signifies that during each round of training, PLG-MOH has already considered the forthcoming data from new classes in the future rounds. Consequently, PLG-MOH can effectively adapt its hashing functions to address the new arriving samples and learn semantic similarity-preserved hash codes for them, meanwhile it can effectively retain the information learned from the old data. Extensive experiments on two public datasets demonstrate that the proposed PLG-MOH achieves better retrieval performance than state-of-the-art baselines on online scenarios.

Recently, large language models (LLMs) have made remarkable progress in table understanding, yet they remain vulnerable to the structural noise (SN) and the textual noise (TN). Existing methods usually employ biased denoising strategies such as structural matching and textual filtering, or overzealous denoising strategies such as introducing supplementary tasks like text-to-SQL and table-to-text to reduce these two types of noise. However, these methods either neglect one type of noise or introduce substantial external noise. Therefore, how to simultaneously mitigate the structural and textual noise without introducing extra noise and improve the performance of LLMs in table understanding is still an unresolved issue. In this paper, we rethink the bottlenecks in table understanding from the perspective of noise reduction and propose a novel dual-denoiser-reasoner model, called TabDDR, for balanced and effective denoising. Specially, our model consists of a structural-and-textual denoiser and a task-adaptive reasoner. The former removes two types of noise via triplet alignment and planning extraction to seek an interpretable balance between breaking structural barriers and preserving structural characteristics, eliminating textual noise and retaining maximal information; the latter ensures a simple but effective reasoning process which can adapt to various downstream tasks. To highlight the presence and impact of the structural and textual noise, we construct the WTQ-SN and WTQ-TN datasets based on the WikiTableQuestion (WTQ) dataset. Extensive experiments on these self-constructed datasets and two other public datasets demonstrate that our proposed method performs better than state-of-the-art baselines.

Task-oriented dialogue (TOD) systems play a vital role in numerous assistance and service scenarios, significantly improving people’s daily lives. Conventionally, a TOD system adheres to a fixed paradigm, where it must first extract user goals and query external databases before it can generate the final response. However, this fixed extract-and-query paradigm is not always optimal for all dialogue turns, which is redundant for the simple turns that do not need external information, and is inadequate for the complex turns that need to interact with the external world multiple times. To address the limitations, in this article, we propose AgentTOD, a novel TOD framework that uses a large language model (LLM) as the intelligent agent to achieve a flexible dialogue paradigm. AgentTOD deprecates the traditional modular architecture (including dialogue state tracking and dialogue policy) by utilizing an LLM as the controller brain to determine when and how to call the provided APIs to obtain external information. It can choose to call APIs any number of times with various parameters until it’s enough to reply to the user. Besides, to train AgentTOD, we construct a large and comprehensive TOD dataset, called TrajsTOD (Trajectories of TODs), which consists of 66k+ user-agent dialogue trajectories converted from eight popular TOD datasets covering 60 domains. TrajsTOD is constructed with minimal dialogue annotations where only the API calling logs are needed and can empower AgentTOD with the general ability to call APIs and generate responses according to the task definition. Extensive experimental results on the MultiWOZ-series and SGD datasets demonstrate AgentTOD has superior performance on TODs as well as a superior adaptability to new task scenarios.

Audio-Visual Semantic Segmentation (AVSS) plays a crucial role in pixel-level multi-modal perception for real-world applications such as robotic navigation and autonomous driving. Existing methods typically rely on global spatio-temporal modules to fuse audio and visual representations, which aids in generating pixel-level semantic masks. However, these approaches often overlook the importance of local spatio-temporal context in understanding semantics, leading to suboptimal performance. This limitation makes it difficult for models to accurately distinguish sound-emitting objects from irrelevant background noise, resulting in erroneous segmentation across the spatio-temporal dimension. To address this issue, we propose the ALOHA framework, which A dapts LO cal spatio-temporal context to en HA nce AVSS. The framework introduces two key components designed to leverage and enhance local spatio-temporal context information: the LOHA adapter and the Selective Context Enhancement (SCE) module. Specifically, the LOHA adapter adaptively captures essential modality information across spatio-temporal dimensions, while implicitly learning fine-grained local context through the local attention mechanism. Furthermore, the SCE module selectively enhances the local context related to the semantics, thereby facilitating the distinction between the sounding object and irrelevant background and improving segmentation accuracy. Moreover, to better adapt to embodied AI systems, our framework utilizes a parameter-shared encoder and applies the adapters in a staged manner. This design significantly reduces the number of trainable parameters, making it more parameter-efficient. Experimental results demonstrate that the proposed framework achieves state-of-the-art performance on the AVSBench-Semantic benchmark dataset and shows competitive results on the AVSBench-Object benchmark, while exhibiting broad adaptability across different visual backbone networks.

Named Entity Recognition (NER), which aims to identify the span and category of entities within text, is a fundamental task in natural language processing. Recent NER approaches have featured pre-trained transformer-based models (e.g., BERT) as a crucial encoding component to achieve state-of-the-art performance. However, due to the length limit for input text, these models typically consider text at the sentence-level and cannot capture the long-range contextual dependency within a document. To address this issue, we propose a novel Span Graph Transformer (SGT) method for document-level NER, which constructs long-range contextual dependencies at both the token and span levels. Specifically, we first retrieve relevant contextual sentences in the document for each target sentence, and jointly encode them by BERT to capture token-level dependencies. Then, our proposed model extracts candidate spans from each sentence and integrates these spans into a document-level span graph, where nested spans within sentences and identical spans across sentences are connected. By leveraging the power of Graph Transformer and well-designed position encoding, our span graph can fully exploit span-level dependencies within the document. Extensive experiments on both resource-rich nested and flat NER datasets, as well as low-resource distantly supervised NER datasets, demonstrate that proposed SGT model achieves better performance than previous state-of-the-art models.

Task-oriented dialogue (TOD) systems are predominantly designed to be composed of several functional modules (e.g.dialogue state tracker, dialogue policy, natural language generation) whether they are pipeline or end-to-end architectures.However, this modular design not only heavily relies on massive fully-annotated data, but also suffers from many intrinsic drawbacks, such as serious error accumulation, poor generalization ability, high customization cost, and low fault tolerance rate.In this paper, we rethink the architecture of the task-oriented dialogue systems and propose a novel fully zeroshot autonomous TOD agent, named AutoTOD, where all the delicate modules in traditional TOD systems are deprecated and all it needs is a general-purpose instruction-following language model (e.g.GPT-4).AutoTOD only leverages a simple instruction schema consisting of the description of tasks and external APIs, and can autonomously decide what to do at each dialogue turn, including asking for information, calling APIs, summarizing API results, and correcting previous mistakes.Moreover, we propose a simulation-based evaluation framework to better validate the abilities of TOD models in real-life scenarios.Extensive experiments conducted on the MultiWOZ and SGD datasets show the superior task completion ability and flexible language skills of AutoTOD. 1

Le Zhuo, Zewen Chi, Minghao Xu, Heyan Huang, Jianan Zhao, Heqi Zheng, Conghui He, Xian-Ling Mao, Wentao Zhang. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Scientific Machine Reading Comprehension (SMRC) aims to facilitate the understanding of scientific texts through human-machine interactions. While existing dataset has significantly contributed to this field, it predominantly focus on single-perspective question-answer pairs, thereby overlooking the inherent variation in comprehension levels among different readers. To address this limitation, we introduce a novel multi-perspective scientific machine reading comprehension dataset, SciMRC, which incorporates perspectives from beginners, students, and experts. Our dataset comprises 741 scientific papers and 6,057 question-answer pairs, with 3,306, 1,800, and 951 pairs corresponding to beginners, students, and experts respectively. Extensive experiments conducted on SciMRC using pre-trained models underscore the importance of considering diverse perspectives in SMRC and highlight the challenging nature of our scientific machine comprehension tasks.

Cross-modal hashing approaches are designed to transform data from disparate modalities into a shared Hamming space while preserving the semantic similarity relationships between the modalities. Generally, most supervised cross-modal hashing methods treat the entire data as input. However, the total information within the data encompasses both vital semantic details and extraneous background noise. Such background information may introduce noise and undermine the preservation of semantic similarity relationships between modalities. To address this issue, we propose a novel Data-Focus Proxy Hashing (DFPH) for cross-modal retrieval, which emphasizes important data information. Our approach begins by training an intermediary network designed to produce a unique proxy hash code specific to each individual category. Then, by incorporating attention mechanisms, and vital textual information, we first design a novel modality-specific hashing network, including an image-focused hash codes generator and a text-focused hash codes generator. Subsequently, we utilize the hashing loss function to train the hashing network under the supervision of these proxy hash codes. Extensive experiments on two benchmarks demonstrate that our proposed DFPH outperforms existing baselines in cross-modal retrieval tasks. Our code is available at https://github.com/JWJ990626/DFPH.

Temporal Knowledge Graph (TKG) is an extension of regular knowledge graph by attaching the time scope. Existing temporal knowledge graph question answering (TKGQA) models solely approach simple questions, owing to the prior assumption that each question only contains a single temporal fact with explicit/implicit temporal constraints. Hence, they perform poorly on questions which own multiple temporal facts. In this paper, we propose Joint Multi Facts Reasoning Network (JMFRN), to jointly reasoning multiple temporal facts for accurately answering complex temporal questions. Specifically, JMFRN first retrieves question-related temporal facts from TKG for each entity of the given complex question. For joint reasoning, we design two different attention (i.e., entity-aware and time-aware) modules, which are suitable for universal settings, to aggregate entities and timestamps information of retrieved facts. Moreover, to filter incorrect type answers, we introduce an additional answer type discrimination task. Extensive experiments demonstrate our proposed method significantly outperforms the state-of-art on the wellknown complex temporal question benchmark TimeQuestions.

Scientific papers of a large scale on the Internet encompass a wealth of data and knowledge, attracting the attention of numerous researchers. To fully utilize these knowledge, Retrieval-Augmented Large Language Models (LLMs) usually leverage large-scale scientific corpus to train and then retrieve relevant passages from external memory to improve generation, which have demonstrated outstanding performance. However, existing methods can only capture one-dimension fragmented textual information without incorporating hierarchical structural knowledge, eg. the deduction relationship of abstract and main body, which makes it difficult to grasp the central thought of papers. To tackle this problem, we propose a hierarchical context augmentation method, which helps Retrieval-Augmented LLMs to autoregressively learn the structure knowledge of scientific papers. Specifically, we utilize the document tree to represent the hierarchical relationship of a paper and enhance the structure information of scientific context from three aspects: scale, format and global information. First, we think each top-bottom path of document tree is a logical independent context, which can be used to largely increase the scale of extracted structural corpus. Second, we propose a novel label-based format to represent the structure of context in textual sequences, unified between training and inference. Third, we introduce the global information of retrieved passages to further enhance the structure of context. Extensive experiments on three scientific tasks show that the proposed method significantly improves the performance of Retrieval-Augmented LLMs on all tasks. Besides, our method achieves start-of-art performance in Question Answer task and outperforms ChatGPT. Moreover, it also brings considerate gains with irrelevant retrieval passages, illustrating its effectiveness on practical application scenarios.

Yuxiang Nie, Heyan Huang, Xian-Ling Mao, Lizi Liao. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Critique ability, i.e., the capability of Large Language Models (LLMs) to identify and rectify flaws in responses, is crucial for their applications in self-improvement and scalable oversight. While numerous studies have been proposed to evaluate critique ability of LLMs, their comprehensiveness and reliability are still limited. To overcome this problem, we introduce CriticEval, a novel benchmark designed to comprehensively and reliably evaluate critique ability of LLMs. Specifically, to ensure the comprehensiveness, CriticEval evaluates critique ability from four dimensions across nine diverse task scenarios. It evaluates both scalar-valued and textual critiques, targeting responses of varying quality. To ensure the reliability, a large number of critiques are annotated to serve as references, enabling GPT-4 to evaluate textual critiques reliably. Extensive evaluations of open-source and closed-source LLMs first validate the reliability of evaluation in CriticEval. Then, experimental results demonstrate the promising potential of open-source LLMs, the effectiveness of critique datasets and several intriguing relationships between the critique ability and some critical factors, including task types, response qualities and critique dimensions.

Topic models that can take advantage of labels are broadly used in identifying interpretable topics from textual data. However, existing topic models tend to merely view labels as names of topic clusters or as categories of texts, thereby neglecting the potential causal relationships between supervised information and latent topics, as well as within these elements themselves. In this paper, we focus on uncovering possible causal relationships both between and within the supervised information and latent topics to better understand the mechanisms behind the emergence of the topics and the labels. To this end, we propose Causal Relationship-Aware Neural Topic Model (CRNTM), a novel neural topic model that can automatically uncover interpretable causal relationships between and within supervised information and latent topics, while concurrently discovering high-quality topics. In CRNTM, both supervised information and latent topics are treated as nodes, with the causal relationships represented as directed edges in a Directed Acyclic Graph (DAG). A Structural Causal Model (SCM) is employed to model the DAG. Experiments are conducted on three public corpora with different types of labels. Experimental results show that the discovered causal relationships are both reliable and interpretable, and the learned topics are of high quality comparing with eight start-of-the-art topic model baselines.

Pretrained language models have achieved great success in a wide range of natural language processing (NLP) problems, because they learn language representations from large-scale text corpora and can adapt to downstream tasks by finetuning them on annotated task data. However, such success relies on both large-scale text and annotated data, so the lack of training data is a major practical problem for many languages, especially low-resource languages. In this paper, we explore whether a pretrained English language model can benefit non-English NLP systems in low-resource scenarios, i.e., with limited text corpora or annotated data. To achieve this, we first propose cross-lingual knowledge transfer methods and then validate our methods in low-resource scenarios. Specifically, our cross-lingual knowledge transfer methods are applied in the training stages of language model pretraining or downstream finetuning. At the two stages, the methods are designed for the transfer of upstream general knowledge or downstream task-specific knowledge, respectively. In the experiments, we perform pretraining and finetuning with limited non-English data to simulate the low-resource scenarios. We evaluate our methods on ten downstream tasks over a wide range of languages, and present systematic comparisons among various knowledge transfer methods. Experimental results show that our methods successfully leverage a pretrained English language model to improve task performance in other languages. Besides, we demonstrate the multilinguality of the English language model in various application scenarios. Our findings imply the possibility to improve low-resource-language NLP systems with large-scale English language models.

Dialogue state tracking plays a key role in tracking user intentions in task-oriented dialogue systems. Traditional dialogue state tracking methods usually rely on selecting slot values from a fixed ontology to represent the dialogue state. In recent years, more flexible open vocabulary based approaches have become the mainstream focus which are mainly divided into two categories: generative methods and span extraction methods. Among them, the span extraction method is favored for its outstanding ability to predict unknown slot values. However, the span extraction method only focuses on the predicted slot values, but ignores other potential slot values in the utterance, which leads to insufficient semantic understanding of the utterance and difficulty in dealing with complex utterance scenarios, such as more or longer unknown slot values. To tackle the above drawbacks, in this paper, we propose a novel scalable dialogue state tracking method, which employs slot tagging to locate all potential slot values in the utterances and jointly learns slot pointers to select the predicted slot value from them. Specifically, our STN4DST (Slot Tagging Navigation for Dialogue State Tracking) model not only adopts the above joint learning strategy, which we call slot tagging navigation, to extract slot values from utterances, but also uses previous dialogue states as dialogue contexts to track the change of slot values, and introduces appendix slot values to predict special slot values that cannot be extracted. Extensive experiments show that in the open vocabulary setting, STN4DST achieves the state-of-the-art joint goal accuracy of 85.4% and 96.5% on Sim-M and Sim-R datasets with a large number of unknown slot values, and is also comparable to other state-of-the-art models in the absence of token-level slot annotations for all potential slot values.

Due to the low storage cost and fast retrieval speed, multi-modal hashing, which maps the instances with different modal data-views into hash codes, has earned increasing research attention. Most existing supervised multi-modal hashing methods exploit the label information to define the similarities between instance pairs to supervise their training process. However, such methods ignore that the transitivity of their defined similarity has been broken in the multi-label scenarios, i.e., the instance$x$is similar to the instance$y$, and the instance$z$is also similar to the instance$y$, but$x$may be not similar to$z$, which will lead to fluctuations in the model optimization process and damage their retrieval performance. For example, in the first batch with instances$x$and$y$but without$z$, the model will be optimized to make the hash codes of$x$and$y$similar to each other; In the second batch with instances$z$and$y$but without$x$, the model will be optimized to make the hash codes of$z$and$y$similar to each other; In the third batch with the instances$x$and$z$but without$y$, the model will be optimized to make the hash codes of$z$and$x$dissimilar to each other, meanwhile in this process, the hash codes of$z$and$x$may be dissimilar to that of$y$which damage the optimizing results of the first two batches. Therefore, we propose a novel Similarity Transitivity Broken-aware Multi-modal Hashing, called STBMH, to solve this problem by adding a novel regularization loss into the original pair-wise loss. For each instance$x$in a training batch, the regularization loss will take all instances in the training set into account. Extensive experiments on four widely used datasets show that the proposed method achieves better performance than the state-of-the-art baselines on multi-modal retrieval task.

Session-based recommendation (SBR) is a challenging task, which aims to predict users’ future interests based on anonymous behavior sequences. Existing methods leverage powerful representation learning approaches to encode sessions into a low-dimensional space. However, despite such achievements, the existing studies focus on the instance-level session learning, while neglecting the group-level users’ preferences (e.g., the common preferences of group users in repeat consumption). To this end, we propose a novelRepeat-awareNeuralMechanism forSession-basedRecommendation (RNMSR). In RNMSR, we propose to learn the user preference from two levels: (i)instance-level, which employs GNNs on a similarity-based item-pairwise session graph to capture the users’ preference in instance-level. (ii)group-level, which converts sessions into group-level behavior patterns to model the group-level users’ preferences. In RNMSR, we combine instance-level and group-level user preference to model the repeat consumption of users, i.e., whether users take repeated consumption and which items are preferred by users. Extensive experiments are conducted on three real-world datasets, i.e., Diginetica, Yoochoose, and Nowplaying, demonstrating that the proposed method consistently achieves state-of-the-art performance in all the tests.

Bundle recommendation aims to recommend a bundle of items for the user to purchase together, for which two scenarios (i.e.Next-bundle recommendation and Within-bundle recommendation) are explored to recommend a specific bundle of items for the user and a specific item to fill the user's current bundle, respectively. Previous works largely model the user's preference with a uniform intent, without considering the diversity of intents when adopting the items within the bundle. In the real scenario of bundle recommendation, user intents modeling actually needs to be considered from three hierarchical levels, for that: a user's intents may be naturally distributed in different bundles (user level), one bundle may contain multiple intents of a user (bundle level), and an item in different bundles may also present different user intents (item level). To this end, we develop a novel model,HierarchicalIntentDisentangleGraphNetworks (HIDGN) for bundle recommendation. HIDGN is capable of capturing the diversity of the user's intent precisely and comprehensively from the hierarchical structure with an cross-task intent contrastive learning, which is unified with the supervised next-/within-bundle recommendation sub-tasks as a multi-task framework. Extensive experiments on three benchmark datasets demonstrate that HIDGN outperforms the state-of-the-art methods by 43.0%, 13.2%, and 73.3%, respectively.

Recent progress in deep learning has continuously improved the accuracy of dialogue response selection. However, in real-world scenarios, the high computation cost forces existing dialogue response selection models to rank only a small number of candidates, recalled by a coarse-grained model, precluding many high-quality candidates. To overcome this problem, we present a novel and efficient response selection model and a set of tailor-designed learning strategies to train it effectively. The proposed model consists of a dense retrieval module and an interaction layer, which could directly select the proper response from a large corpus. We conduct re-rank and full-rank evaluations on widely used benchmarks to evaluate our proposed model. Extensive experimental results demonstrate that our proposed model notably outperforms the state-of-the-art baselines on both re-rank and full-rank evaluations. Moreover, human evaluation results show that the response quality could be improved further by enlarging the candidate pool with nonparallel corpora. In addition, we also release high-quality benchmarks that are carefully annotated for more accurate dialogue response selection evaluation. All source codes, datasets, model parameters, and other related resources have been publicly available. 1

Coarse-grained response selection is a fundamental and essential subsystem for the widely used retrieval-based chatbots, aiming to recall a coarse-grained candidate set from a large-scale dataset. The dense retrieval technique has recently been proven very effective in building such a subsystem. However, dialogue dense retrieval models face two problems in real scenarios: (1) the multi-turn dialogue history is re-computed in each turn, leading to inefficient inference; (2) the index storage of the offline index is enormous, significantly increasing the deployment cost. To address these problems, we propose an efficient coarse-grained response selection subsystem consisting of two novel methods. Specifically, to address the first problem, we propose the H ierarchical D ense R etrieval. It caches rich multi-vector representations of the dialogue history and only encodes the latest user’s utterance, leading to better inference efficiency. Then, to address the second problem, we design the D eep S emantic H ashing to reduce the index storage while effectively saving its recall accuracy notably. Extensive experimental results prove the advantages of the two proposed methods over previous works. Specifically, with the limited performance loss, our proposed coarse-grained response selection model achieves over 5x FLOPs speedup and over 192x storage compression ratio. Moreover, our source codes have been publicly released. 1

Aspect Sentiment Triplet Extraction (ASTE) has become an emerging task in sentiment analysis research, aiming to extract triplets of the aspect term, its corresponding opinion term, and its associated sentiment polarity from a given sentence. Recently, many neural networks based models with different tagging schemes have been proposed, but almost all of them have their limitations: heavily relying on 1) prior assumption that each word is only associated with a single role (e.g., aspect term, or opinion term, etc. ) and 2) word-level interactions and treating each opinion/aspect as a set of independent words. Hence, they perform poorly on the complex ASTE task, such as a word associated with multiple roles or an aspect/opinion term with multiple words. Hence, we propose a novel approach, Span TAgging and Greedy infErence (STAGE), to extract sentiment triplets in span-level, where each span may consist of multiple words and play different roles simultaneously. To this end, this paper formulates the ASTE task as a multi-class span classification problem. Specifically, STAGE generates more accurate aspect sentiment triplet extractions via exploring span-level information and constraints, which consists of two components, namely, span tagging scheme and greedy inference strategy. The former tag all possible candidate spans based on a newly-defined tagging set. The latter retrieves the aspect/opinion term with the maximum length from the candidate sentiment snippet to output sentiment triplets. Furthermore, we propose a simple but effective model based on the STAGE, which outperforms the state-of-the-arts by a large margin on four widely-used datasets. Moreover, our STAGE can be easily generalized to other pair/triplet extraction tasks, which also demonstrates the superiority of the proposed scheme STAGE.

Open-retrieval conversational machine reading comprehension (OCMRC) simulates reallife conversational interaction scenes.Machines are required to make a decision of Yes/No/Inquire or generate a follow-up question when the decision is Inquire based on retrieved rule texts, user scenario, user question and dialogue history.Recent studies try to reduce the information gap between decision-making and question generation, in order to improve the performance of generation.However, the information gap still persists because these methods are still limited in pipeline framework, where decision-making and question generation are performed separately, making it hard to share the entailment reasoning used in decision-making across all stages.To tackle the above problem, we propose a novel one-stage end-to-end framework, called Entailment Fused-T5 (EFT), to bridge the information gap between decisionmaking and question generation in a global understanding manner.The extensive experimental results demonstrate that our proposed framework achieves new state-of-the-art performance on the OR-ShARC benchmark.Our model and code are publicly available 1 .

Conversational recommender systems (CRS) aim to timely trace the dynamic interests of users through dialogues and generate relevant responses for item recommendations.Recently, various external knowledge bases (especially knowledge graphs) are incorporated into CRS to enhance the understanding of conversation contexts.However, recent reasoning-based models heavily rely on simplified structures such as linear structures or fixed-hierarchical structures for causality reasoning, hence they cannot fully figure out sophisticated relationships among utterances with external knowledge.To address this, we propose a novel Treestructure Reasoning schEmA named TREA.TREA constructs a multi-hierarchical scalable tree as the reasoning structure to clarify the causal relationships between mentioned entities, and fully utilizes historical conversations to generate more reasonable and suitable responses for recommended results.Extensive experiments on two public CRS datasets have demonstrated the effectiveness of our approach.Our

Video captioning aims at producing textual descriptions for the given video. Benefiting from the self-attention mechanism for capturing long-distance dependencies between video patches and language sentences, the fully Transformer-based models achieve promising performance recently. However, due to continuous temporal information, there exists a large amount of redundant and unimportant visual content. Indiscriminate use of video patches results in expensive computation and inefficient use of resources. To tackle this issue, we propose Ada-SwinBERT, a novel approach that adaptively selects salient video tokens to achieve a balance between efficiency and performance for video captioning. Moreover, we devise a training strategy with online self-distillation to make up for the information loss caused by discarding video tokens. Video-text alignment knowledge distilled from the teacher leads to a robust training process. By pruning 78.1% input tokens hierarchically, our approach greatly reduces 62.0% FLOPs compared with the base model while achieving competitive performance with SOTA methods.

Generalization beyond in-domain experience to out-of-distribution data is of paramount significance in the AI domain. Of late, state-of-the-art Visual Question Answering (VQA) models have shown impressive performance on in-domain data, partially due to the language prior bias which, however, hinders the generalization ability in practice. This paper attempts to provide new insights into the influence of language modality on VQA performance from an empirical study perspective. To achieve this, we conducted a series of experiments on six models. The results of these experiments revealed that, 1) apart from prior bias caused by question types, there is a notable influence of postfix-related bias in inducing biases, and 2) training VQA models with word-sequence-related variant questions demonstrated improved performance on the out-of-distribution benchmark, and the LXMERT even achieved a 10-point gain without adopting any debiasing methods. We delved into the underlying reasons behind these experimental results and put forward some simple proposals to reduce the models' dependency on language priors. The experimental results demonstrated the effectiveness of our proposed method in improving performance on the out-of-distribution benchmark, VQA-CPv2. We hope this study can inspire novel insights for future research on designing bias-reduction approaches.

Conversational recommendation systems (CRS) aim to timely and proactively acquire user dynamic preferred attributes through conversations for item recommendation. In each turn of CRS, there naturally have two decision-making processes with different roles that influence each other: 1) director, which is to select the follow-up option (i.e., ask or recommend) that is more effective for reducing the action space and acquiring user preferences; and 2) actor, which is to accordingly choose primitive actions (i.e., asked attribute or recommended item) to estimate the effectiveness of the director’s option. However, existing methods heavily rely on a unified decision-making module or heuristic rules, while neglecting to distinguish the roles of different decision procedures, as well as the mutual influences between them. To address this, we propose a novel Director-Actor Hierarchical Conversational Recommender (DAHCR), where the director selects the most effective option, followed by the actor accordingly choosing primitive actions that satisfy user preferences. Specifically, we develop a dynamic hypergraph to model user preferences and introduce an intrinsic motivation to train from weak supervision over the director. Finally, to alleviate the bad effect of model bias on the mutual influence between the director and actor, we model the director’s option by sampling from a categorical distribution. Extensive experiments demonstrate that DAHCR outperforms state-of-the-art methods.

Recently, numerous proxy hash code based methods, which sufficiently exploit the label information of data to supervise the training of hashing models, have been proposed. Although these methods have made impressive progress, their generating processes of proxy hash codes are based only on the class information of the dataset or labels of data but do not take the data themselves into account. Therefore, these methods will probably generate some inappropriate proxy hash codes, thus damaging the retrieval performance of the hash models. To solve the aforementioned problem, we propose a novel Data-Aware Proxy Hashing for cross-modal retrieval, called DAPH. Specifically, our proposed method first train a data-aware proxy network that takes the data points, label vectors of data, and the class vectors of the dataset as inputs to generate class-based data-aware proxy hash codes, label-fused image-aware proxy hash codes and label-fused text-aware proxy hash codes. Then, we propose a novel hash loss that exploits the three types of data-aware proxy hash codes to supervise the training of modality-specific hashing networks. After training, DAPH is able to generate discriminate hash codes with the semantic information preserved adequately. Extensive experiments on three benchmark datasets show that the proposed DAPH outperforms the state-of-the-art baselines in cross-modal retrieval tasks.

Conversational recommendation systems (CRS) aim to interactively acquire user preferences and accordingly recommend items to users. Accurately learning the dynamic user preferences is of crucial importance for CRS. Previous works learn the user preferences with pairwise relations from the interactive conversation and item knowledge, while largely ignoring the fact that factors for a relationship in CRS are multiplex. Specifically, the user likes/dislikes the items that satisfy some attributes (Like/Dislike view). Moreover social influence is another important factor that affects user preference towards the item (Social view), while is largely ignored by previous works in CRS. The user preferences from these three views are inherently different but also correlated as a whole. The user preferences from the same views should be more similar than that from different views. The user preferences from Like View should be similar to Social View while different from Dislike View. To this end, we propose a novel model, namely Multi-view Hypergraph Contrastive Policy Learning (MHCPL). Specifically, MHCPL timely chooses useful social information according to the interactive history and builds a dynamic hypergraph with three types of multiplex relations from different views. The multiplex relations in each view are successively connected according to their generation order in the interactive conversation. A hierarchical hypergraph neural network is proposed to learn user preferences by integrating information of the graphical and sequential structure from the dynamic hypergraph. A cross-view contrastive learning module is proposed to maintain the inherent characteristics and the correlations of user preferences from different views. Extensive experiments conducted on benchmark datasets demonstrate that MHCPL outperforms the state-of-the-art methods.

Recently, to improve the unsupervised image retrieval performance, plenty of unsupervised hashing methods have been proposed by designing a semantic similarity matrix, which is based on the similarities between image features extracted by a pre-trained CNN model. However, most of these methods tend to ignore high-level abstract semantic concepts contained in images. Intuitively, concepts play an important role in calculating the similarity among images. In real-world scenarios, each image is associated with some concepts, and the similarity between two images will be larger if they share more identical concepts. Inspired by the above intuition, in this work, we propose a novel Unsupervised Hashing with Semantic Concept Mining, called UHSCM, which leverages a VLP model to construct a high-quality similarity matrix. Specifically, a set of randomly chosen concepts is first collected. Then, by employing a vision-language pretraining (VLP) model with the prompt engineering which has shown strong power in visual representation learning, the set of concepts is denoised according to the training images. Next, the proposed method UHSCM applies the VLP model with prompting again to mine the concept distribution of each image and construct a high-quality semantic similarity matrix based on the mined concept distributions. Finally, with the semantic similarity matrix as guiding information, a novel hashing loss with a modified contrastive loss based regularization item is proposed to optimize the hashing network. Extensive experiments on three benchmark datasets show that the proposed method outperforms the state-of-the-art baselines in the image retrieval task.

Neural variational inference-based topic modeling has gained great success in mining abstract topics from documents. However, these topic models usually mainly focus on optimizing the topic proportions for documents, while the quality and the internal construction of topics are usually neglected. Specifically, these models lack the guarantee that semantically related words are supposed to be assigned to the same topic and are difficult to ensure the interpretability of topics. Moreover, many topical words recur frequently in the top words of different topics, which makes the learned topics semantically redundant and similar, and of little significance for further study. To solve the above problems, we propose a novel neural topic model called Neural Variational Gaussian Mixture Topic Model (NVGMTM). We use Gaussian distribution to depict the semantic relevance between words in the topics. Each topic in NVGMTM is considered as a multivariate Gaussian distribution over words in the word-embedding space. Thus, semantically related words share similar probabilities in each topic, which makes the topics more coherent and interpretable. Experimental results on two public corpora show the proposed model outperforms the state-of-the-art baselines.

With the explosively increasing of academic papers, efficient academic document retrieval is becoming an essential requirement for large-scale information retrieval systems. Inspired by the success of deep semantic hashing in normal document retrieval, deep semantic hashing is a promising approach for academic document retrieval by mapping academic documents into efficient hash codes. However, for academic document retrieval, the existing deep semantic hashing methods suffer from following two problems: (1) they cannot differentiate the importance of different field labels; (2) they cannot plenty utilize the structure information in paper citations. To address these problems, we propose a novel Large-scale Academic deep Semantic Hashing, called LASH. Specifically, LASH first treats paper citations as a citation network, and then employs a multi-input variational deep autoencoder to directly encode both structure information of the citation network and semantic information of academic documents into unified hash codes. Moreover, a weighted percentage similarity is designed to measure the importance of different field labels, which is a linear combination of Jaccard and Cosine similarity. Supervised by the similarity, the learned unified hash codes can further preserve the importance of different field labels. Extensive experiments show LASH significantly outperforms state-of-the-art baselines over proposed three real-world large-scale academic datasets.

Document hashing is a powerful paradigm for document retrieval, which maps high-dimensional documents to compact hashing codes with preserving the similarity of original data. While fairly successful, the existing document hashing methods do not consider the relevance relationship among different documents from a category and the hierarchical relationship among categories. Intuitively, the intra-category relevance connects related concepts among different documents, which can supplement the omitted information for each document; meanwhile the hierarchical categories can help to identify whether mistakes occur in leaf categories or parent categories, which can be used to reduce the mistakes occurring in parent categories that are often more serious. Inspired by above intuitions, we propose a novel \textbf{I}ntra-category aware \textbf{H}ierarchical supervised \textbf{D}ocument \textbf{H}ashing, called IHDH. Specifically, IHDH is a binary autoencoder architecture equipped with two novel components: intra-category component and hierarchy component. The intra-category component exploits the difference among latent semantic representations of different documents from a category to supplement the omitted information for each document. The hierarchy component utilizes the hierarchical structure to transform the probabilities of leaf categories into the probabilities of parent categories by union operation, and then gives a further parent-level penalty to reduce the mistakes occurring in parent categories.

Due to the high retrieval efficiency and low storage cost for cross-modal search tasks, cross-modal hashing methods have attracted considerable attention from the researchers. For the supervised cross-modal hashing methods, how to make the learned hash codes sufficiently preserve semantic information contained in the label of datapoints is the key to further enhance the retrieval performance. Hence, almost all supervised cross-modal hashing methods usually depend on defining similarities between datapoints with the label information to guide the hashing model learning fully or partly. However, the defined similarity between datapoints can only capture the label information of datapoints partially and misses abundant semantic information, which then hinders the further improvement of retrieval performance. Thus, in this paper, different from previous works, we propose a novel cross-modal hashing method without defining the similarity between datapoints, called Deep Cross-modal Proxy Hashing (DCPH). Specifically, DCPH first trains a proxy hashing network to transform each category information of a dataset into a semantic discriminative hash code, called proxy hash code. Each proxy hash code can preserve the semantic information of its corresponding category well. Next, without defining the similarity between datapoints to supervise the training process of the modality-specific hashing networks, we propose a novelmargin-dynamic-softmax lossto directly utilize the proxy hashing codes as supervised information. Finally, by minimizing the novelmargin-dynamic-softmax loss, the modality-specific hashing networks can be trained to generate hash codes that can simultaneously preserve the cross-modal similarity and abundant semantic information well. Extensive experiments on three benchmark datasets show that the proposed method outperforms the state-of-the-art baselines in the cross-modal retrieval tasks.

Visual relationship understanding plays an indispensable role in grounded language tasks like visual question answering (VQA), which often requires precisely reasoning about relations among objects depicted in the given question. However, prior works generally suffer from the deficiencies as follows, (1)spatial-relation inference ambiguity, it is challenging to accurately estimate the distance of a pair of visual objects in 2D space if there is a visual-overlap between their 2D bounding-boxes, and (2) language-visual relational alignment missing, it is insufficient to generate a high-quality answer to the question if there is a lack of alignment in the language-visual relations of objects during fusion, even using a powerful fusion model like Transformer. To this end, we first model the spatial relation of a pair of objects in 3D space by augmenting the original 2D bounding-box with 1D depth information, and then propose a novel model namedDepth-awareSemanticGuided RelationalAttention Network (DSGANet), to explicitly exploit the formed 3D spatial relations of objects in an intra-/inter-modality manner for precise relational alignment. Extensive experiments conducted on the benchmarks (VQA v2.0 and GQA) demonstrate DSGANet achieves competitive performance compared to pretrained and non-pretrained models, such as 72.7% vs. 74.6% based on the learned grid features on VQA v2.0.

Cross-modal hashing has been widely used in multimedia retrieval tasks due to its fast retrieval speed and low storage cost. Recently, many deep unsupervised cross-modal hashing methods have been proposed to deal the unlabeled datasets. These methods usually construct an instance similarity matrix by fusing the image and text modality-specific similarity matrices as the guiding information to train the hashing networks. However, most of them directly use cosine similarities between the bag-of-words (BoW) vectors of text datapoints to define the text modality-specific similarity matrix, which fails to mine the semantic similarity information contained in the text modal datapoints and leads to the poor quality of the instance similarity matrix. To tackle the aforementioned problem, in this paper, we propose a novel Unsupervised Cross-modal Hashing via Semantic Text Mining, called UCHSTM. Specifically, UCHSTM first mines the correlations between the words of text datapoints. Then, UCHSTM constructs the text modality-specific similarity matrix for the training instances based on the mined correlations between their words. Next, UCHSTM fuses the image and text modality-specific similarity matrices as the final instance similarity matrix to guide the training of hashing model. Furthermore, during the process of training the hashing networks, a novel self-redefined-similarity loss is proposed to further correct some wrong defined similarities in the constructed instance similarity matrix, thereby further enhancing the retrieval performance. Extensive experiments on two widely used datasets show that the proposed UCHSTM outperforms state-of-the-art baselines on cross-modal retrieval tasks. We provide our source codes at:https://github.com/rongchengtu1/UCHTIM.

Relation ties, defined as the correlation and mutual exclusion between different relations, are critical for distant supervised relation extraction. Previous studies usually obtain this property by greedily learning the local connections between relations. However, they are essentially limited because of failing to capture the global topology structure of relation ties and may easily fall into a locally optimal solution. To address this issue, we propose a novel force-directed graph to comprehensively learn relation ties. Specifically, we first construct a graph according to the global co-occurrence of all relations. Then, we borrow the idea of Coulomb’s law from physics and introduce the concept of attractive force and repulsive force into this graph to learn correlation and mutual exclusion between relations. Finally, the obtained relation representations are applied as an inter-dependent relation classifier. Extensive experimental results demonstrate that our method is capable of modeling global correlation and mutual exclusion between relations, and outperforms the state-of-the-art baselines. In addition, the proposed force-directed graph can be used as a module to augment existing relation extraction systems and improve their performance.

Joint entity and relation extraction is an essential task in natural language processing and knowledge graph construction. Existing approaches usually decompose the joint extraction task into several basic modules or processing steps to make it easy to conduct. However, such a paradigm ignores the fact that the three elements of a triple are interdependent and indivisible. Therefore, previous joint methods suffer from the problems of cascading errors and redundant information. To address these issues, in this paper, we propose a novel joint entity and relation extraction model, named OneRel, which casts joint extraction as a fine-grained triple classification problem. Specifically, our model consists of a scoring-based classifier and a relation-specific horns tagging strategy. The former evaluates whether a token pair and a relation belong to a factual triple. The latter ensures a simple but effective decoding process. Extensive experimental results on two widely used datasets demonstrate that the proposed method performs better than the state-of-the-art baselines, and delivers consistent performance gain on complex scenarios of various overlapping patterns and multiple triples.

Bundle recommendation aims to recommend the user a bundle of items as a whole. Previous models capture user’s preferences on both items and the association of items. Nevertheless, they usually neglect the diversity of user’s intents on adopting items and fail to disentangle user’s intents in representations. In the real scenario of bundle recommendation, a user’s intent may be naturally distributed in the different bundles of that user (Global view). And a bundle may contain multiple intents of a user (Local view). Each view has its advantages for intent disentangling: 1) In the global view, more items are involved to present each intent, which can demonstrate the user’s preference under each intent more clearly. 2) The local view can reveal the association between items under each intent since the items within the same bundle are highly correlated to each other. To this end, in this paper we propose a novel model named Multi-view Intent Disentangle Graph Networks (MIDGN), which is capable of precisely and comprehensively capturing the diversity of user intent and items’ associations at the finer granularity. Specifically, MIDGN disentangles user’s intents from two different perspectives, respectively: 1) taking the Global view, MIDGN disentangles the user’s intent coupled with inter-bundle items; 2) taking the Local view, MIDGN disentangles the user’s intent coupled with items within each bundle. Meanwhile, we compare user’s intents disentangled from different views by a contrast method to improve the learned intents. Extensive experiments are conducted on two benchmark datasets and MIDGN outperforms the state-of-the-art methods by over 10.7% and 26.8%, respectively.

Zewen Chi, Shaohan Huang, Li Dong, Shuming Ma, Bo Zheng, Saksham Singhal, Payal Bajaj, Xia Song, Xian-Ling Mao, Heyan Huang, Furu Wei. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022.

Heqi Zheng, Xiao Zhang, Zewen Chi, Heyan Huang, Yan Tan, Tian Lan, Wei Wei, Xian-Ling Mao. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022.

Endowing chatbots with a consistent personality plays a vital role for agents to deliver human-like interactions. However, existing personalized approaches commonly generate responses in light of static predefined personas depicted with textual description, which may severely restrict the interactivity of human and the chatbot, especially when the agent needs to answer the query excluded in the predefined personas, which is so-called out-of-predefined persona problem (named OOP for simplicity). To alleviate the problem, in this paper we propose a novel retrieval-to-prediction paradigm consisting of two subcomponents, namely, (1) Persona Retrieval Model (PRM), it retrieves a persona from a global collection based on a Natural Language Inference (NLI) model, the inferred persona is consistent with the predefined personas; and (2) Posterior-scored Transformer (PS-Transformer), it adopts a persona posterior distribution that further considers the actual personas used in the ground response, maximally mitigating the gap between training and inferring. Furthermore, we present a dataset called IT-ConvAI2 that first highlights the OOP problem in personalized dialogue. Extensive experiments on both IT-ConvAI2 and ConvAI2 demonstrate that our proposed model yields considerable improvements in both automatic metrics and human evaluations.

Sequential recommendation (SR) aims to predict the subsequent behaviors of users by understanding their successive historical behaviors. Recently, some methods for SR are devoted to alleviating the data sparsity problem (i.e., limited supervised signals for training), which take account of contrastive learning to incorporate self-supervised signals into SR. Despite their achievements, it is far from enough to learn informative user/item embeddings due to the inadequacy modeling of complex collaborative information and co-action information, such as user-item relation, user-user relation, and item-item relation. In this paper, we study the problem of SR and propose a novel multi-level contrastive learning framework for sequential recommendation, named MCLSR. Different from the previous contrastive learning-based methods for SR, MCLSR learns the representations of users and items through a cross-view contrastive learning paradigm from four specific views at two different levels (i.e., interest- and feature-level). Specifically, the interest-level contrastive mechanism jointly learns the collaborative information with the sequential transition patterns, and the feature-level contrastive mechanism re-observes the relation between users and items via capturing the co-action information (i.e., co-occurrence). Extensive experiments on four real-world datasets show that the proposed MCLSR outperforms the state-of-the-art methods consistently.

It is the most important way for researchers to acquire academic progress via reading scientific papers, most of which are in PDF format. However, existing PDF Readers like Adobe Acrobat Reader and Foxit PDF Reader are usually only for reading by rendering PDF files as a whole, and do not consider the multi-granularity content understanding of a paper itself. Specifically, taking a paper as a basic and separate unit, existing PDF Readers cannot access extended information about the paper, such as corresponding videos, blogs and codes. Meanwhile, they cannot understand the academic content of a paper, such as terms, authors, and citations. To solve these problems, we introduce Hammer PDF, an intelligent PDF Reader for scientific papers. Apart from basic reading functions, Hammer PDF has the following four innovative features: (1) information extraction ability, which can locate and mark spans like terms and other entities; (2) information extension ability, which can present relevant academic content of a paper, such as citations, references, codes, videos, blogs, etc; (3) built-in Hammer Scholar, an academic search engine based on academic information collected from major academic databases; (4) built-in Q&A bot, which can find helpful conference information. The proposed Hammer PDF Reader can help researchers, especially those studying computer science, to improve the efficiency and experience of reading scientific papers. We have released Hammer PDF, available at https://pdf.hammerscholar.net/face.

Incorporating Knowledge Graphs (KG) into recommeder system as side information has attracted considerable attention. Recently, the technical trend of Knowledge-aware Recommendation (KGR) is to develop end-to-end models based on graph neural networks (GNNs). However, the extremely sparse user-item interactions significantly degrade the performance of the GNN-based models, from the following aspects: 1) the sparse interaction, itself, means inadequate supervision signals and limits the supervised GNN-based models; 2) the combination of sparse interactions (CF part) and redundant KG facts (KG part) further results in an unbalanced information utilization. Besides, the GNN paradigm aggregates local neighbors for node representation learning, while ignoring the non-local KG facts and making the knowledge extraction insufficient. Inspired by the recent success of contrastive learning in mining supervised signals from data itself, in this paper, we focus on exploring contrastive learning in KGR and propose a novel multi-level interactive contrastive learning mechanism, to alleviate the aforementioned challenges. Different from traditional contrastive learning methods which contrast nodes of two generated graph views, interactive contrastive mechanism conducts layer-wise self-supervised learning by contrasting layers of different parts within graphs, which is also an "interaction" action. Specifically, we first construct local and non-local graphs for user/item in KG, exploring more KG facts for KGR. Then an intra-graph level interactive contrastive learning is performed within each local/non-local graph, which contrasts layers of the CF and KG parts, for more consistent information leveraging. Besides, an inter-graph level interactive contrastive learning is performed between the local and non-local graphs, for sufficiently and coherently extracting non-local KG signals. Extensive experiments conducted on three benchmark datasets show the superior performance of our proposed method over the state-of-the-arts. The implementations are available at: https://github.com/CCIIPLab/KGIC.

Currently, human-bot symbiosis dialog systems, e.g. pre- and after-sales in E-commerce, are ubiquitous, and the dialog routing component is essential to improve the overall efficiency, reduce human resource cost and increase user experience. To satisfy this requirement, existing methods are mostly heuristic and cannot obtain high-quality performance. In this paper, we investigate the important problem by thoroughly mining both the data-to-task and task-to-task knowledge among various kinds of dialog data. To achieve the above target, we propose a comprehensive and general solution with multi-task learning framework, specifically including a novel dialog encoder and two tailored gated mechanism modules. The proposed Gated Mechanism enhanced Multi-task Model (G3M) can play the role of hierarchical information filtering and is non-invasive to the existing dialog systems. Experiments on two datasets collected from the real world demonstrate our method’s effectiveness and the results achieve the state-of-the-art performance by relatively increasing 8.7%/11.8% on RMSE metric and 2.2%/4.4% on F1 metric.

Unsupervised question answering is an attractive task due to its independence on labeled data. Previous works usually make use of heuristic rules as well as pre-trained models to construct data and train QA models. However, most of these works regard named entity (NE) as the only answer type, which ignores the high diversity of answers in the real world. To tackle this problem, we propose a novel unsupervised method by diversifying answers, named DiverseQA. Specifically, the proposed method is composed of three modules: data construction, data augmentation and denoising filter. Firstly, the data construction module extends the extracted named entity into a longer sentence constituent as the new answer span to construct a QA dataset with diverse answers. Secondly, the data augmentation module adopts an answer-type dependent data augmentation process via adversarial training in the embedding level. Thirdly, the denoising filter module is designed to alleviate the noise in the constructed data. Extensive experiments show that the proposed method outperforms previous unsupervised models on five benchmark datasets, including SQuADv1.1, NewsQA, TriviaQA, BioASQ, and DuoRC. Besides, the proposed method shows strong performance in the few-shot learning setting.

Conversational machine reading comprehension (CMRC) aims to assist computers to understand an natural language text and thereafter engage in a multi-turn conversation to answer questions related to the text. Existing methods typically require three steps: (1) decision making based on entailment reasoning; (2) span extraction if required by the above decision; (3) question rephrasing based on the extracted span. However, for nearly all these methods, the span extraction and question rephrasing steps cannot fully exploit the fine-grained entailment reasoning information in decision making step because of their relative independence, which will further enlarge the information gap between decision making and question phrasing. Thus, to tackle this problem, we propose a novel end-to-end framework for conversational machine reading comprehension based on shared parameter mechanism, called entailment reasoning T5 (ET5). Despite the lightweight of our proposed framework, experimental results show that the proposed ET5 achieves new state-of-the-art results on the ShARC leaderboard with the BLEU-4 score of 55.2. Our model and code are publicly available.

Long document question answering is a challenging task due to its demands for complex reasoning over long text.Previous works usually take long documents as non-structured flat texts or only consider the local structure in long documents.However, these methods usually ignore the global structure of the long document, which is essential for long-range understanding.To tackle this problem, we propose Compressive Graph Selector Network (CGSN) to capture the global structure in a compressive and iterative manner.The proposed model mainly focuses on the evidence selection phase of long document question answering.Specifically, it consists of three modules: local graph network, global graph network and evidence memory network.Firstly, the local graph network builds the graph structure of the chunked segment in token, sentence, paragraph and segment levels to capture the short-term dependency of the text.Secondly, the global graph network selectively receives the information of each level from the local graph, compresses them into the global graph nodes and applies graph attention to the global graph nodes to build the long-range reasoning over the entire text in an iterative way.Thirdly, the evidence memory network is designed to alleviate the redundancy problem in the evidence selection by saving the selected result in the previous steps.Extensive experiments show that the proposed model outperforms previous methods on two datasets.

Extracting relational triples from unstructured text is an essential task in natural language processing and knowledge graph construction. Existing approaches usually contain two fundamental steps: (1) finding the boundary positions of head and tail entities; (2) concatenating specific tokens to form triples. However, nearly all previous methods suffer from the problem of error accumulation, i.e., the boundary recognition error of each entity in step (1) will be accumulated into the final combined triples. To solve the problem, in this paper, we introduce a fresh perspective to revisit the triple extraction task and propose a simple but effective model, named DirectRel. Specifically, the proposed model first generates candidate entities through enumerating token sequences in a sentence, and then transforms the triple extraction task into a linking problem on a ``head -> tail" bipartite graph. By doing so, all triples can be directly extracted in only one step. Extensive experimental results on two widely used datasets demonstrate that the proposed model performs better than the state-of-the-art baselines.

Recently, the research of dialogue systems has been widely concerned, especially task-oriented dialogue systems, which have received increased attention due to their wide application prospect. As a core component, dialogue state tracking (DST) plays a key role in task-oriented dialogue systems, and its function is to parse natural language dialogues into dialogue state formed by slot-value pairs. It is well known that dialogue state tracking has been well studied and explored on current benchmark datasets such as the MultiWOZ. However, almost all current research completely ignores the user negative feedback utterances that exist in real-life conversations when a system error occurs, which often contains user-provided corrective information for the system error. Obviously, user negative feedback utterances can be used to correct the inevitable errors in automatic speech recognition and model generalization. Thus, in this paper, we will explore the role of negative feedback utterances in dialogue state tracking in detail through simulated negative feedback utterances. Specifically, due to the lack of dataset involving negative feedback utterances, first, we have to define the schema of user negative feedback utterances and propose a joint modeling method for feedback utterance generation and filtering. Then, we explore three aspects of interaction mechanism that should be considered in real-life conversations involving negative feedback utterances and propose evaluation metrics related to negative feedback utterances. Finally, on WOZ2.0 and MultiWOZ2.1 datasets, by constructing simulated negative feedback utterances in training and testing, we not only verify the important role of negative feedback utterances in dialogue state tracking, but also analyze the advantages and disadvantages of different interaction mechanisms involving negative feedback utterances, lighting future research on negative feedback utterances.

Sparse mixture of experts provides larger model capacity while requiring a constant computational overhead. It employs the routing mechanism to distribute input tokens to the best-matched experts according to their hidden representations. However, learning such a routing mechanism encourages token clustering around expert centroids, implying a trend toward representation collapse. In this work, we propose to estimate the routing scores between tokens and experts on a low-dimensional hypersphere. We conduct extensive experiments on cross-lingual language model pre-training and fine-tuning on downstream tasks. Experimental results across seven multilingual benchmarks show that our method achieves consistent gains. We also present a comprehensive analysis on the representation and routing behaviors of our models. Our method alleviates the representation collapse issue and achieves more consistent routing than the baseline mixture-of-experts methods.

Knowledge graph (KG) plays an increasingly important role in recommender systems. Recently, graph neural networks (GNNs) based model has gradually become the theme of knowledge-aware recommendation (KGR). However, there is a natural deficiency for GNN-based KGR models, that is, the sparse supervised signal problem, which may make their actual performance drop to some extent. Inspired by the recent success of contrastive learning in mining supervised signals from data itself, in this paper, we focus on exploring the contrastive learning in KG-aware recommendation and propose a novel multi-level cross-view contrastive learning mechanism, named MCCLK. Different from traditional contrastive learning methods which generate two graph views by uniform data augmentation schemes such as corruption or dropping, we comprehensively consider three different graph views for KG-aware recommendation, including global-level structural view, local-level collaborative and semantic views. Specifically, we consider the user-item graph as a collaborative view, the item-entity graph as a semantic view, and the user-item-entity graph as a structural view. MCCLK hence performs contrastive learning across three views on both local and global levels, mining comprehensive graph feature and structure information in a self-supervised manner. Besides, in semantic view, a k-Nearest-Neighbor (k NN) item-item semantic graph construction module is proposed, to capture the important item-item semantic relation which is usually ignored by previous work. Extensive experiments conducted on three benchmark datasets show the superior performance of our proposed method over the state-of-the-arts. The implementations are available at: https://github.com/CCIIPLab/MCCLK.

Due to their high retrieval efficiency and low storage cost, cross-modal hashing methods have attracted considerable attention. Generally, compared with shallow cross-modal hashing methods, deep cross-modal hashing methods can achieve a more satisfactory performance by integrating feature learning and hash codes optimizing into a same framework. However, most existing deep cross-modal hashing methods either cannot learn a unified hash code for the two correlated data-points of different modalities in a database instance or cannot guide the learning of unified hash codes by the feedback of hashing function learning procedure, to enhance the retrieval accuracy. To address the issues above, in this paper, we propose a novel end-to-end Deep Cross-Modal Hashing with Hashing Functions and Unified Hash Codes Jointly Learning (DCHUC). Specifically, by an iterative optimization algorithm, DCHUC jointly learns unified hash codes for image-text pairs in a database and a pair of hash functions for unseen query image-text pairs. With the iterative optimization algorithm, the learned unified hash codes can be used to guide the hashing function learning procedure; Meanwhile, the learned hashing functions can feedback to guide the unified hash codes optimizing procedure. Extensive experiments on three public datasets demonstrate that the proposed method outperforms the state-of-the-art cross-modal hashing methods.

Zewen Chi, Li Dong, Bo Zheng, Shaohan Huang, Xian-Ling Mao, Heyan Huang, Furu Wei. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Puhai Yang, Heyan Huang, Xian-Ling Mao. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Zewen Chi, Li Dong, Shuming Ma, Shaohan Huang, Saksham Singhal, Xian-Ling Mao, Heyan Huang, Xia Song, Furu Wei. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. 2021.

Recently, deep Hamming hashing methods have been proposed for Hamming space retrieval which enables constant-time search by hash table lookups instead of linear scan. When carrying out Hamming space retrieval, for each query datapoint, there is a Hamming ball centered on the query datapoint, and only the datapoints within the Hamming ball are returned as the relevant ones, while those beyond are discarded directly. Thus, to further enhance the retrieval performance, it is a key point for the Hamming hashing methods to decrease the dissimilar datapoints within the Hamming ball. However, nearly all existing Hamming hashing methods cannot effectively penalize the dissimilar pairs within the Hamming ball to push them out. To tackle this problem, in this paper, we propose a novel Weighted Gaussian Loss based Hamming Hashing, called WGLHH, which introduces a weighted Gaussian loss to optimize hashing model. Specifically, the weighted Gaussian loss consists of three parts: a novel Gaussian-distribution based loss, a novel badly-trained-pair attention mechanism and a quantization loss. The Gaussian-distribution based loss is proposed to effectively penalize the dissimilar pairs within the Hamming ball. The badly-trained-pair attention mechanism is proposed to assign a weight for each data pair, which puts more weight on data pairs whose corresponding hash codes cannot preserve original similarity well, and less on those having already handled well. The quantization loss is used to reduce the quantization error. By incorporating the three parts, the proposed weighted Gaussian loss will penalize significantly on the dissimilar pairs within the Hamming ball to generate more compact hashing codes. Extensive experiments on two benchmark datasets show that the proposed method outperforms the state-of-the-art baselines in image retrieval task.

Zewen Chi, Li Dong, Furu Wei, Nan Yang, Saksham Singhal, Wenhui Wang, Xia Song, Xian-Ling Mao, Heyan Huang, Ming Zhou. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Recently, deep supervised hashing methods have shown state-of-the-art performance by integrating feature learning and hash codes learning into an end-to-end network to generate high-quality hash codes. However, it is still a challenge to learn discriminative hash codes for preserving the label information of images efficiently. To overcome this difficulty, in this paper, we propose a novel Partial-Softmax Loss based Deep Hashing, called PSLDH, to generate high-quality hash codes. Specifically, PSLDH first trains a category hashing network to generate a discriminative hash code for each category, and the hash code will preserve semantic information of the corresponding category well. Then, instead of defining the similarity between datapairs using their corresponding label vectors, we directly use the learned hash codes of categories to supervise the learning process of image hashing network, and a novel Partial-SoftMax loss is proposed to optimize the image hashing network. By minimizing the novel Partial-SoftMax loss, the learned hash codes can preserve the label information of images sufficiently. Extensive experiments on three benchmark datasets show that the proposed method outperforms the state-of-the-art baselines in image retrieval task.

The consistency of a response to a given post at the semantic level and emotional level is essential for a dialogue system to deliver humanlike interactions. However, this challenge is not well addressed in the literature, since most of the approaches neglect the emotional information conveyed by a post while generating responses. This article addresses this problem and proposes a unified end-to-end neural architecture, which is capable of simultaneously encoding the semantics and the emotions in a post and leveraging target information to generate more intelligent responses with appropriately expressed emotions. Extensive experiments on real-world data demonstrate that the proposed method outperforms the state-of-the-art methods in terms of both content coherence and emotion appropriateness.

In this work we focus on transferring supervision signals of natural language generation (NLG) tasks between multiple languages. We propose to pretrain the encoder and the decoder of a sequence-to-sequence model under both monolingual and cross-lingual settings. The pre-training objective encourages the model to represent different languages in the shared space, so that we can conduct zero-shot cross-lingual transfer. After the pre-training procedure, we use monolingual data to fine-tune the pre-trained model on downstream NLG tasks. Then the sequence-to-sequence model trained in a single language can be directly evaluated beyond that language (i.e., accepting multi-lingual input and producing multi-lingual output). Experimental results on question generation and abstractive summarization show that our model outperforms the machine-translation-based pipeline methods for zero-shot cross-lingual generation. Moreover, cross-lingual transfer improves NLG performance of low-resource languages by leveraging rich-resource language data. Our implementation and data are available at https://github.com/CZWin32768/xnlg.

The noisy labeling problem has been one of the major obstacles for distant supervised relation extraction. Existing approaches usually consider that the noisy sentences are useless and will harm the model's performance. Therefore, they mainly alleviate this problem by reducing the influence of noisy sentences, such as applying bag-level selective attention or removing noisy sentences from sentence-bags. However, the underlying cause of the noisy labeling problem is not the lack of useful information, but the missing relation labels. Intuitively, if we can allocate credible labels for noisy sentences, they will be transformed into useful training data and benefit the model's performance. Thus, in this paper, we propose a novel method for distant supervised relation extraction, which employs unsupervised deep clustering to generate reliable labels for noisy sentences. Specifically, our model contains three modules: a sentence encoder, a noise detector and a label generator. The sentence encoder is used to obtain feature representations. The noise detector detects noisy sentences from sentence-bags, and the label generator produces high-confidence relation labels for noisy sentences. Extensive experimental results demonstrate that our model outperforms the state-of-the-art baselines on a popular benchmark dataset, and can indeed alleviate the noisy labeling problem.

Most of the unsupervised hashing methods usually map images into semantic similarity-preserving hash codes by constructing local semantic similarity structure as guiding information, i.e., treating each point similar to its k nearest neighbours. However, for an image, some of its k nearest neighbours may be dissimilar to it, i.e., they are noisy datapoints which will damage the retrieval performance. Thus, to tackle this problem, in this paper, we propose a novel deep unsupervised hashing method, called MLS3RDUH, which can reduce the noisy datapoints to further enhance retrieval performance. Specifically, the proposed method first defines a novel similarity matrix by utilising the intrinsic manifold structure in feature space and the cosine similarity of datapoints to reconstruct the local semantic similarity structure. Then a novel log-cosh hashing loss function is used to optimize the hashing network to generate compact hash codes by incorporating the defined similarity as guiding information. Extensive experiments on three public datasets show that the proposed method outperforms the state-of-the-art baselines.

Session-based recommendation (SBR) is a challenging task, which aims at recommending items based on anonymous behavior sequences. Almost all the existing solutions for SBR model user preference only based on the current session without exploiting the other sessions, which may contain both relevant and irrelevant item-transitions to the current session. This paper proposes a novel approach, called Global Context Enhanced Graph Neural Networks (GCE-GNN) to exploit item transitions over all sessions in a more subtle manner for better inferring the user preference of the current session. Specifically, GCE-GNN learns two levels of item embeddings from session graph and global graph, respectively: (i) Session graph, which is to learn the session-level item embedding by modeling pairwise item-transitions within the current session; and (ii) Global graph, which is to learn the global-level item embedding by modeling pairwise item-transitions over all sessions. In GCE-GNN, we propose a novel global-level item representation learning layer, which employs a session-aware attention mechanism to recursively incorporate the neighbors' embeddings of each node on the global graph. We also design a session-level item representation learning layer, which employs a GNN on the session graph to learn session-level item embeddings within the current session. Moreover, GCE-GNN aggregates the learnt item representations in the two levels with a soft attention mechanism. Experiments on three benchmark datasets demonstrate that GCE-GNN outperforms the state-of-the-art methods consistently.

Open-domain generative dialogue systems have attracted considerable attention over the past few years. Currently, how to automatically evaluate them is still a big challenge. As far as we know, there are three kinds of automatic evaluations for open-domain generative dialogue systems: (1) Word-overlap-based metrics; (2) Embedding-based metrics; (3) Learning-based metrics. Due to the lack of systematic comparison, it is not clear which kind of metrics is more effective. In this article, we first measure systematically all kinds of metrics to check which kind is best. Extensive experiments demonstrate that learning-based metrics are the most effective evaluation metrics for open-domain generative dialogue systems. Moreover, we observe that nearly all learning-based metrics depend on the negative sampling mechanism, which obtains extremely imbalanced and low-quality samples to train a score model. To address this issue, we propose a novel learning-based metric that significantly improves the correlation with human judgments by using augmented PO sitive samples and valuable NE gative samples, called PONE. Extensive experiments demonstrate that PONE significantly outperforms the state-of-the-art learning-based evaluation method. Besides, we have publicly released the codes of our proposed metric and state-of-the-art baselines. 1

The consistency of a response to a given post at semantic-level and emotional-level is essential for a dialogue system to deliver human-like interactions. However, this challenge is not well addressed in the literature, since most of the approaches neglect the emotional information conveyed by a post while generating responses. This article addresses this problem by proposing a unified end-to-end neural architecture, which is capable of simultaneously encoding the semantics and the emotions in a post for generating more intelligent responses with appropriately expressed emotions. Extensive experiments on real-world data demonstrate that the proposed method outperforms the state-of-the-art methods in terms of both content coherence and emotion appropriateness.

As a critical component of Spoken Dialog System (SDS), spoken language understanding (SLU) attracts a lot of attention, especially for methods based on unaligned data. Recently, a new approach has been proposed that utilizes the hierarchical relationship between act-slot-value triples. However, it ignores the transfer of internal information which may record the intermediate information of the upper level and contribute to the prediction of the lower level. So, we propose a novel streamlined decoding structure with attention mechanism, which uses three successively connected RNN to decode act, slot and value respectively. On the first Chinese Audio-Textual Spoken Language Understanding Challenge (CATSLU), our model exceeds state-of-the-art model on an unaligned multi-turn task-oriented Chinese spoken dialogue dataset provided by the contest.

Recently, similarity-preserving hashing methods have been extensively studied for large-scale image retrieval. Compared with unsupervised hashing, supervised hashing methods for labeled data have usually better performance by utilizing semantic label information. Intuitively, for unlabeled data, it will improve the performance of unsupervised hashing methods if we can first mine some supervised semantic 'label information' from unlabeled data and then incorporate the 'label information' into the training process. Thus, in this paper, we propose a novel Object Detection based Deep Unsupervised Hashing method (ODDUH). Specifically, a pre-trained object detection model is utilized to mining supervised 'label information', which is used to guide the learning process to generate high-quality hash codes. Extensive experiments on two public datasets demonstrate that the proposed method outperforms the state-of-the-art unsupervised hashing methods in the image retrieval task.

Learning an efficient manager of dialogue agent from data with little manual intervention is important, especially for goal-oriented dialogues.However, existing methods either take too many manual efforts (e.g.reinforcement learning methods) or cannot guarantee the dialogue efficiency (e.g.sequence-to-sequence methods).In this paper, we address this problem by proposing a novel end-to-end learning model to train a dialogue agent that can look ahead for several future turns and generate an optimal response to make the dialogue efficient.Our method is data-driven and does not require too much manual work for intervention during system design.We evaluate our method on two datasets of different scenarios and the experimental results demonstrate the efficiency of our model.

Understanding unstructured texts is an essential skill for human beings as it enables knowledge acquisition. Although understanding unstructured texts is easy for we human beings with good education, it is a great challenge for machines. Recently, with the rapid development of artificial intelligence techniques, researchers put efforts to teach machines to understand texts and justify the educated machines by letting them solve the questions upon the given unstructured texts, inspired by the reading comprehension test as we humans do. However, feature effectiveness with respect to different questions significantly hinders the performance of answer selection, because different questions may focus on various aspects of the given text and answer candidates. To solve this problem, we propose a question-oriented feature attention (QFA) mechanism, which learns to weight different engineering features according to the given question, so that important features with respect to the specific question is emphasized accordingly. Experiments on MCTest dataset have well-validated the effectiveness of the proposed method. Additionally, the proposed QFA is applicable to various IR tasks, such as question answering and answer selection. We have verified the applicability on a crawled community-based question-answering dataset.

Recently, hashing methods have been widely used in large-scale image retrieval. However, most existing supervised hashing methods do not consider the hierarchical relation of labels,which means that they ignored the rich semantic information stored in the hierarchy. Moreover, most of previous works treat each bit in a hash code equally, which does not meet the scenario of hierarchical labeled data. To tackle the aforementioned problems, in this paper, we propose a novel deep hashing method, called supervised hierarchical deep hashing (SHDH), to perform hash code learning for hierarchical labeled data. Specifically, we define a novel similarity formula for hierarchical labeled data by weighting each level, and design a deep neural network to obtain a hash code for each data point. Extensive experiments on two real-world public datasets show that the proposed method outperforms the state-of-the-art baselines in the image retrieval task.

To compare the similarity of probability distributions, the information-theoretically motivated metrics like Kullback-Leibler divergence (KL) and Jensen-Shannon divergence (JSD) are often more reasonable compared with metrics for vectors like Euclidean and angular distance. However, existing locality-sensitive hashing (LSH) algorithms cannot support the information-theoretically motivated metrics for probability distributions. In this paper, we first introduce a new approximation formula for S2JSD-distance, and then propose a novel LSH scheme adapted to S2JSD-distance for approximate nearest neighbors search in high-dimensional probability distributions. We define the specific hashing functions, and prove their local-sensitivity. Furthermore, extensive empirical evaluations well illustrate the effectiveness of the proposed hashing schema on six public image datasets and two text datasets, in terms of mean Average Precision, Precision@N and Precision-Recall curve.

Sentence auto-completion is an important feature that saves users many keystrokes in typing the entire sentence by providing suggestions as they type. Despite its value, the existing sentence auto-completion methods, such as query completion models, can hardly be applied to solving the object completion problem in sentences with the form of (subject, verb, object), due to the complex natural language description and the data deficiency problem. Towards this goal, we treat an SVO sentence as a three-element triple (subject, sentence pattern, object), and cast the sentence object completion problem as an element inference problem. These elements in all triples are encoded into a unified low-dimensional embedding space by our proposed TRANSFER model, which leverages the external knowledge base to strengthen the representation learning performance. With such representations, we can provide reliable candidates for the desired missing element by a linear model. Extensive experiments on a real-world dataset have well-validated our model. Meanwhile, we have successfully applied our proposed model to factoid question answering systems for answer candidate selection, which further demonstrates the applicability of the TRANSFER model.

Recently, topic modeling has been widely applied in data mining due to its powerful ability. A common, major challenge in applying such topic models to other tasks is to accurately interpret the meaning of each topic. Topic labeling, as a major interpreting method, has attracted significant attention recently. However, most of previous works only focus on the effectiveness of topic labeling, and less attention has been paid to quickly creating good topic descriptors; meanwhile, it’s hard to assign labels for new emerging topics by using most of existing methods. To solve the problems above, in this paper, we propose a novel fast topic labeling framework that casts the labeling problem as a k-nearest neighbor (KNN) search problem in a probability vector set. Our experimental results show that the proposed sequential interleaving method based on locality sensitive hashing (LSH) technology is efficient in boosting the comparison speed among probability distributions, and the proposed framework can generate meaningful labels to interpret topics, including new emerging topics.

Recently, authority ranking has received increasing interests in both academia and industry, and it is applicable to many problems such as discovering influential nodes and building recommendation systems. Various graph-based ranking approaches like PageRank have been used to rank authors and papers separately in homogeneous networks. In this paper, we take venue information into consideration and propose a novel graph-based ranking framework, Tri-Rank, to co-rank authors, papers and venues simultaneously in heterogeneous networks. This approach is a flexible framework and it ranks authors, papers and venues iteratively in a mutually reinforcing way to achieve a more synthetic, fair ranking result. We conduct extensive experiments using the data collected from ACM Digital Library. The experimental results show that Tri-Rank is more effective and efficient than the state-of-the-art baselines including PageRank, HITS and Co-Rank in ranking authors. The papers and venues ranked by Tri-Rank also demonstrate that Tri-Rank is rational.

Lots of document collections are well organized in hierarchical structure, and such structure can help users browse and understand these collections. Meanwhile, there are a large number of plain document collections loosely organized, and it is difficult for users to understand them effectively. In this paper we study how to automatically integrate latent topics in a plain collection with the topics in a hierarchical structured collection. We propose to use semi-supervised topic modeling to solve the problem in a principled way. The experiments show that the proposed method can generate both meaningful latent topics and expand high quality hierarchical topic structures.

Recently, statistical topic modeling has been widely applied in text mining and knowledge management due to its powerful ability. A topic, as a probability distribution over words, is usually difficult to be understood. A common, major challenge in applying such topic models to other knowledge management problem is to accurately interpret the meaning of each topic. Topic labeling, as a major interpreting method, has attracted significant attention recently. However, previous works simply treat topics individually without considering the hierarchical relation among topics, and less attention has been paid to creating a good hierarchical topic descriptors for a hierarchy of topics. In this paper, we propose two effective algorithms that automatically assign concise labels to each topic in a hierarchy by exploiting sibling and parent-child relations among topics. The experimental results show that the inter-topic relation is effective in boosting topic labeling accuracy and the proposed algorithms can generate meaningful topic labels that are useful for interpreting the hierarchical topics.