Next POI recommendation aids users in predicting their destinations of interest and plays an increasingly vital role in location-based social services. Recent works focus on analyzing both long-term and short-term interests in POI recommendation to gain a deeper understanding of user profiles. However, these methods for modeling long-term user’s sequences primarily rely on the Transformer model, which functions as a low-pass filter, often leading to the loss of high-frequency information. Additionally, long-term and short-term sequences are typically modeled independently, with short-term sequences often defined solely by the most recent check-ins, overlooking their interactions and dependencies. Therefore, we propose Enhancing Long-and Short-Term Representations for Next POI Recommendations via Frequency and Hierarchical Contrastive Learning (FHCRec). FHCRec captures both high-frequency and low-frequency information in long-term sequences to model richer long-term user’s preference representations. Moreover, it harnesses the characteristics of the short-term subsequences embedded within long-term sequences to enhance short-term preference characterization via local and global hierarchical contrastive learning, resulting in more personalized short-term preferences. The enhanced long-term and short-term preferences are integrated to improve model recommendation performance. Extensive experiments on three real-world datasets demonstrate the effectiveness of our method.

The Knowledge Graph logical reasoning task faces a dual challenge of insufficient semantic coverage from type information and missing structural information from relations. Although type annotations provide semantic priors for entities, their coarse-grained features cannot comprehensively characterize entity attributes; conversely, relational structure can enhance semantic representation, but the incompleteness of edges in real-world graphs limits modeling when relying on a single information source. To address these issues, we propose MRCLQR (Multi-information Relation Constraint-based Logical Query Reasoning), a framework with three core innovations: (1) an Information Semantic Alignment module based on contrastive learning, which achieves cross-modal semantic collaboration via entity-type-structure pairing; (2) a Constraint-aware Relation Encoding method that decomposes relation semantics into domain aggregation features, relation ontology semantics, and range constraint features; and (3) Neural-Symbolic Operators guided by domain constraints, which narrow the reasoning space through a constraint-aware attention mechanism. Experiments on FB15k, FB15k-237, and NELL-995 demonstrate that MRCLQR achieves average MRR scores of 35.8%, 16.2%, and 19.6%, respectively improving over the strongest baselines by 0.5%, 0.2%, and 0.2% --- and exhibits an 8.0% average gain on complex queries involving negation. Ablation studies validate the effectiveness of multi-source collaboration and the curriculum learning strategy. This work offers a novel paradigm for heterogeneous knowledge fusion and logical query reasoning.

Nowadays, misinformation articles, especially multimodal ones, are widely spread on social media platforms and cause serious negative effects. To control their propagation, Multimodal Misinformation Detection (MMD) becomes an active topic in the community to automatically identify misinformation. Previous MMD methods focus on supervising detectors by collecting offline data. However, in real-world scenarios, new events always continually emerge, making MMD models trained on offline data consistently outdated and ineffective. To address this issue, training MMD models under online data streams is an alternative, inducing an emerging task named continual MMD. Unfortunately, it is hindered by two major challenges. First, training on new data consistently decreases the detection performance on past data, named past knowledge forgetting. Second, the social environment constantly evolves over time, affecting the generalization on future data. To alleviate these challenges, we propose to remember past knowledge by isolating interference between event-specific parameters with a Dirichlet process-based mixture-of-expert structure, and anticipate future environmental distributions by learning a continuous-time dynamics model. Accordingly, we induce a new continual MMD method DAEDCMD. Extensive experiments demonstrate that DAEDCMD can consistently and significantly outperform the compared methods, including six MMD baselines and three continual learning methods.

Positive and Unlabeled (PU) learning is a special case of binary classification with weak supervision, where only positive labeled and unlabeled data are available. Previous studies suggest several specific risk estimators of PU learning such as non-negative PU (nnPU), which are unbiased and consistent with the expected risk of supervised binary classification. In nnPU, the negative-class empirical risk is estimated by positive labeled and unlabeled data with a non-negativity constraint. However, its negative-class empirical risk estimator approaches 0, so the negative class is over-played, resulting in imbalanced error rates between positive and negative classes. To solve this problem, we suppose that the expected risks of the positive-class and negative-class should be close. Accordingly, we constrain that the negative-class empirical risk estimator is lower bounded by the positive-class empirical risk, instead of 0; and also incorporate an explicit equality constraint between them. we suggest a risk estimator of PU learning that balances positive and negative classification error rates, named $\mathrm{D{\small C-PU} }$, and suggest an efficient training method for $\mathrm{D{\small C-PU} }$ based on the augmented Lagrange multiplier framework. We theoretically analyze the estimation error of $\mathrm{D{\small C-PU} }$ and empirically validate that $\mathrm{D{\small C-PU} }$ achieves higher accuracy and converges more stable than other risk estimators of PU learning. Additionally, $\mathrm{D{\small C-PU} }$ also performs competitive accuracy performance with practical PU learning methods.

Trajectory metric learning, which supports the trajectory similarity search, is one of the most fundamental tasks in spatial-temporal data analysis. However, existing trajectory metric learning methods rely on massive labels of pairwise trajectory distance, and thus cannot be applied to few-shot scenarios frequently occurring in real-world applications. Though performance drops caused by insufficient labels can be alleviated by knowledge distillation, we demonstrate that they cannot be directly applied to few-shot trajectory metric learning due to the domain shift problem. To this end, this paper proposes invariant and relaxed learning enhanced knowledge distillation method TMLKD for few-shot trajectory metric learning, such that domain-invariant representation and rank knowledge can be distilled. Specifically, in the representation learning phase, it first employs an adversarial sub-network to distinguish domain-specific and domain-invariant information, so as to distill transferable representation knowledge from teacher models. To mitigate the few-shot problem in student model training, we further enrich sparse labels of the target domain by utilizing the rank knowledge revealed in teachers' predictions. Particularly, TMLKD employs a list-wise learning-to-rank approach to learn the relaxed trajectory ranking orders instead of focusing on all the samples inefficiently. Finally, to guide accurate distillation, we adaptively assign reliability of teacher prediction by utilizing the ground-truth labels, to avoid misleading the student model with low-quality teacher predictions. Extensive experiments on three real-world datasets demonstrate the superiority of our model.

Knowledge-grounded dialogue (KGD) learns to generate an informative response based on a given dialogue context and external knowledge (e.g., knowledge graphs; KGs). Recently, the emergence of large language models (LLMs) and pre-training techniques has brought great success to knowledge-grounded dialogue. However, when building KGD systems in real applications, there are various real-world noises that are inevitable to face. For example, the dialogue context might involve perturbations such as misspellings and abbreviations. In addition, KGs typically suffer from incompletion and also might contain erroneous and outdated facts. Such real-world noises pose a challenge to the robustness of KGD systems and hinder their applications in the real world. In this paper, we propose an entity-based contrastive learning framework for improving the robustness of KGD. Specifically, we make use of the entity information in a KGD sample to create both its positive and negative samples which involve semantic-irrelevant and semantic-relevant perturbations, respectively. The contrastive learning framework ensures the KGD model is aware of these two types of perturbations, thus could generate informative responses with the potentially noisy inputs in real applications. Experimental results on three widely-used benchmark datasets show that our method achieves new state-of-the-art performance in terms of automatic evaluation scores, verifying its effectiveness and potentiality. Furthermore, we show that our method is able to generate better responses than comparison models in both the noisy and the few-shot settings.

The automatic generation of radiological imaging reports aims to produce accurate and coherent clinical descriptions based on X-ray images. This facilitates clinicians in completing the arduous task of report writing and advances clinical automation. The primary challenge in radiological imaging report generation lies in accurately capturing and describing abnormal regions in the images under data bias conditions, resulting in the generation of lengthy texts containing image details. Existing methods mostly rely on prior knowledge such as medical knowledge graphs, corpora, and image databases to assist models in generating more precise textual descriptions. However, these methods still struggle to identify rare anomalies in the images. To address this issue, we propose a two-stage training model, named CLR2G, based on cross-modal contrastive learning. This model delegates the task of capturing anomalies, particularly those challenging for the generative model trained with cross-entropy loss under data bias conditions, to a specialized abnormality capture component. Specifically, we employ a semantic matching loss function to train additional abnormal image and text encoders through cross-modal contrastive learning, facilitating the capture of 13 common anomalies. We utilize the anomalous image features, text features and their confidence probabilities as a posteriori knowledge to help the model generate accurate image reports. Experimental results demonstrate the state-of-the-art performance of our method on two widely used public datasets, IU-Xray and MIMIC-CXR.

Sequential Recommendation (SR) has received increasing attention due to its ability to capture user dynamic preferences. Recently, Contrastive Learning (CL) provides an effective approach for sequential recommendation by learning invariance from different views of an input. However, most existing data or model augmentation methods may destroy semantic sequential interaction characteristics and often rely on the hand-crafted property of their contrastive view-generation strategies. In this paper, we propose a Meta-optimized Seq2Seq Generator and Contrastive Learning (Meta-SGCL) for sequential recommendation, which applies the meta-optimized two-step training strategy to adaptive generate contrastive views. Specifically, Meta-SGCL first introduces a simple yet effective augmentation method called Sequence-to-Sequence (Seq2Seq) generator, which treats the Variational AutoEncoders (VAE) as the view generator and can constitute contrastive views while preserving the original sequence's semantics. Next, the model employs a meta-optimized two-step training strategy, which aims to adaptively generate contrastive views without relying on manually designed view-generation techniques. Finally, we evaluate our proposed method Meta-SGCL using three public real-world datasets. Compared with the state-of-the-art methods, our experimental results demonstrate the effectiveness of our model and the code is available.11https.//anonymous.4open.science/status/Meta-SGCL-05B5

Graph Collaborative Filtering (GCF) is designed to leverage high-order connectivity in user-item graphs, thereby significantly enhancing recommendation performance. Recent advancements have seen the integration of contrastive learning into GCF as a strategy to mitigate the challenges of data sparsity. This approach involves creating contrastive views through augmentations, followed by the generation of self-supervised signals. These signals are produced by maximizing the mutual information between the contrastive views. While this method has proven effective, we argue that current CL-based GCF models are still limited to current augmentation techniques. Existing data augmentation or noise perturbation may destroy the structural and semantic features of the original data and node attribute information is not considered. To tackle the above limitations, we propose a Meta-optimized Structure and Semantic Contrastive Learning for Graph Collaborative Filtering, named Meta-SSCL, which utilizes graph structure information and semantic information contrastive learning for recommendation. Specifically, we first model the structural and node semantic information representations with LightGCN and vanilla attention mechanism, respectively. Then consider the structural and semantic information as two contrastive views for recommendation. Next, the meta-optimized two-step training strategy generates adaptive contrastive views. Finally, we fuse structural and semantic representations for recommendation. Extensive experiments on real-world datasets demonstrate that Meta-SSCL consistently outperforms state-of-the-art sequential recommendation methods. The code is available1.

Multi-modal data have recently been utilized to improve the performance of knowledge graph completion (KGC), attracting widespread research interest. However, they have been ignored in few-shot knowledge graph completion (FKGC), which aims to discover potential facts involving unseen relations that only appear in few-shot triples. The most relevant FKGC study simply concatenates various modal features, but the performance is still limited due to the following problems: (1) lack of exploiting significant multi-modal features in neighborhoods, and (2) ineffectively modeling inter-modal interactions in a few-shot setting. To tackle these problems, we propose a novel relational learning model entitled MMSN (Multi-Modal Siamese Network) for few-shot knowledge graph completion, which is composed of the following two primary modules: the Siamese multi-modal neighbor encoder (SMNE) and the meta-learning multi-modal knowledge representation decoder (MKRD). The module SMNE is developed to encode diverse modalities of neighbors by a Siamese attention network, fuse multi-modal information through a gating fusion network, and learn effective relational embeddings using an aggregator. The module MKRD is introduced to handle inter-modal interactions between multiple modalities and train the proposed model in a few-shot scenario. Extensive experiments demonstrate that our proposed model MMSN outperforms the state-of-the-art FKGC models, including uni-modal and multi-modal models, on two real-world few-shot multi-modal datasets.

Extracting entities and relations from text is a significant task of information extraction. Existing extraction models often straightforwardly produce their confident prediction results without any reconsideration or double-checking, resulting in avoidable mistakes and sub-optimal performance. In this paper, we propose a novel coarse-to-fine extraction framework, which first extracts high-potential relations as well as entities via knowledge distillation, and then rechecks the predictions via handcrafted natural language inference (NLI) task in a fine-grained manner. Specifically, based on the knowledge distillation mechanism, we train multiple teacher models iteratively through an adaptive loss function for making one teacher concentrate more on the data that others are incompetent for. Then, these complementary teacher models are utilized to provide valuable soft-label information for training a considerate student model, enabling it to generate reliable preliminary predictions. Further, these generated potential relations and entities are formulated as hypotheses, together with the original sentences as premises, serving as the input for an NLI model. Considering the linguistic diversity of relational expression, we automatically generate various semantic templates for hypotheses through an$\mathcal{N}$-gram mining strategy. Moreover, due to the existence of multi-fact sentences, a relation-guided Gaussian attention is designed to reduce the gap between the single-relation hypothesis and the multi-relation premise. To implement efficient training, we also develop several ways to generate high-quality negative samples, which help the NLI model learn to identify errors. Experimental results show that the proposed method is effective and outperforms other strong baselines on public benchmarks.

Self-supervised contrastive learning has well advanced the development of bundle recommendation. However, self-supervised contrastive learning may misclassify some positive samples as negative samples, resulting in suboptimal models. Therefore, the industry has proposed supervised contrastive learning to alleviate this drawback. Inspired by this idea, we seek a more elegant contrastive learning paradigm in the field of recommendation, so we propose a dual-supervised contrastive learning for bundle recommendation (DSCBR), which integrates supervised and self-supervised contrastive learning to exploit the full potential of contrastive learning. Specifically, we first construct self-supervised contrastive learning between the two different views (bundle view and item view), which encourages the alignment of the two separately learned views and boosts the effectiveness of learned representations. Second, we use the interaction information to construct supervised contrastive learning and leverage the bundle–bundle cooccurrence graph for further enhancement. By introducing supervised contrastive learning, our model explicitly models user and bundle proximity in different views, improving the model’s robustness and generalization. Finally, we jointly perform self-supervised and supervised contrastive learning across multiple views. Extensive experiments on three public datasets demonstrate the effectiveness of our model.

Document-level Relation Extraction (RE) is a promising task aiming at identifying relations of multiple entity pairs in a document. Compared with the sentence-level counterpart, it has raised two significant challenges: a) In most cases, a relational fact can be adequately expressed via a small subset of sentences from the document, namely evidence. But the traditional method cannot model such strong semantic correlations between evidence sentences that collaborate to describe a specific relation; b) The data of this task is extremely long-tail in terms of too many NA instances and imbalanced relational types. Such data can mislead the tail prediction bias to the head categories in the RE model. In this paper, we present a novelEvidence reasoning andCurriculum learning method forDocRE(DRE-EC) to address these challenges. Particularly, we first formulate evidence extraction as a sequential decision problem through a crafted reinforcement learning mechanism with an efficient path searching strategy to reduce the action space. Providing the evidence for each entity pair as a customized-filtered document in advance helps infer the relations better. To address the long-tail issue, we further develop a hybrid curriculum learning method at the NA-level (NC) and relation-level (RC) with our customized difficulty measure score. In NC, the NA samples are scheduled in an easy-to-hard scheme and gradually added, resulting in the data distribution from ideal and balanced to real and unbalanced. In RC, the scheme is switched into hard-to-easy to enhance the hard and tail samples. In addition, we propose a new Equalization adaptive Focal Loss(EFLoss) that can adjust to the changing data distribution and focus more on the tail categories. We conduct various experiments on two document-level RE benchmarks and achieve a remarkable improvement over previous competitive baselines. Furthermore, we provide detailed analyses of the advantages and effectiveness of our method.

Semi-Supervised Relation Extraction aims at learning well-performed RE models with limited labeled and large-scale unlabeled data. Existing methods mainly suffer from semantic drift and insufficient supervision, which severely limit the performance. To address these problems, recent work tends to design dual modules to work cooperatively for mutual enhancement. However, the consensus of two modules greatly restricts the model from exploring diverse relation expressions in unlabeled set, which hinders the performance as well as model generalization. To tackle this problem, in this paper, we propose a novel competition-based method AdvSRE. We set up a challenging minimax game on unlabeled data between two modules, Generator and Discriminator, and assign them with conflicting objectives. During the competition game, one module may find any possible chance to beat the other, which develops two modules' abilities until relation expressions cannot be further explored. To exploit label information, Discriminator is further asked to predict specific relation for each sentence. Experiment results on two benchmarks show new state-of-the-art performance over baselines, demonstrating the effectiveness of proposed AdvSRE.

Jiaan Wang, Fandong Meng, Duo Zheng, Yunlong Liang, Zhixu Li, Jianfeng Qu, Jie Zhou. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Multi-modal knowledge graphs (MKGs) include not only the relation triplets, but also related multi-modal auxiliary data (i.e., texts and images), which enhance the diversity of knowledge. However, the natural incompleteness has significantly hindered the applications of MKGs. To tackle the problem, existing studies employ the embedding-based reasoning models to infer the missing knowledge after fusing the multi-modal features. However, the reasoning performance of these methods is limited due to the following problems: (1) ineffective fusion of multi-modal auxiliary features; (2) lack of complex reasoning ability as well as inability to conduct the multi-hop reasoning which is able to infer more missing knowledge. To overcome these problems, we propose a novel model entitled MMKGR (Multi-hop Multi-modal Knowledge Graph Reasoning). Specifically, the model contains the following two components: (1) a unified gate-attention network which is designed to generate effective multi-modal complementary features through sufficient attention interaction and noise reduction; (2) a complementary feature-aware reinforcement learning method which is proposed to predict missing elements by performing the multi-hop reasoning process, based on the features obtained in component (1). The experimental results demonstrate that MMKGR outperforms the state-of-the-art approaches in the MKG reasoning task.

The self-attention mechanism, which equips with a strong capability of modeling long-range dependencies, is one of the extensively used techniques in the sequential recommendation field. However, many recent studies represent that current self-attention based models are low-pass filters and are inadequate to capture high-frequency information. Furthermore, since the items in the user behaviors are intertwined with each other, these models are incomplete to distinguish the inherent periodicity obscured in the time domain. In this work, we shift the perspective to the frequency domain, and propose a novel Frequency Enhanced Hybrid Attention Network for Sequential Recommendation, namely FEARec. In this model, we firstly improve the original time domain self-attention in the frequency domain with a ramp structure to make both low-frequency and high-frequency information could be explicitly learned in our approach. Moreover, we additionally design a similar attention mechanism via auto-correlation in the frequency domain to capture the periodic characteristics and fuse the time and frequency level attention in a union model. Finally, both contrastive learning and frequency regularization are utilized to ensure that multiple views are aligned in both the time domain and frequency domain. Extensive experiments conducted on four widely used benchmark datasets demonstrate that the proposed model performs significantly better than the state-of-the-art approaches.

Cross-Lingual Summarization (CLS) aims at generating summaries in one language for the given documents in another language. CLS has attracted wide research attention due to its practical significance in the multi-lingual world. Though great contributions have been made, existing CLS works typically focus on short documents, such as news and guides. Different from these short texts, long documents such as academic articles usually discuss complicated subjects and consist of thousands of words, making them non-trivial to process and summarize. To promote CLS research on long documents, we construct Perseus, the first long-document CLS dataset which collects about 94K Chinese scientific documents paired with English summaries. The average length of documents in Perseus is more than 2000 tokens. As a preliminary study on long-document CLS, we build and evaluate various CLS baselines, including pipeline and end-to-end methods. Experimental results on Perseus show the superiority of the end-to-end baseline, which performs the best among all methods. Furthermore, to provide a deeper understanding, we manually analyze the model outputs and discuss specific challenges faced by current approaches. We hope that our work could benchmark long-document CLS and benefit future studies.

Distant supervision is an efficient way to generate large-scale training data for relation extraction without human efforts. However, the accompanying challenges have been plaguing the advance of the extractor: (1) the automatically annotated labels for training data contain much noisy data; (2) the annotations, based on bag-level (cluster of sentences) instead of sentence-level (single sentence), are too coarse to train an accurate extractor; (3) hetergeneous sentences are hard for a denoising model to capture the underlying commonness among valid relational expressions. To address these issues, we bulid a novel sentence representation and craft reinforcement learning to select the expressive sentence for each relation mentioned in a bag. More specifically, we introduce entity-free sentence pattern incorporated with attentive type information. Furthermore, multiple interactions between entity-specific and entity-free representation are proposed to generate complementary sentence features (for challenge 3). Then we design a fine-grained reward function, and model the sentence selection process as an auction where different relations for a bag need to compete together to achieve the possession of a specific sentence based on its expressiveness(for challenge 1 and 2). The experimental results on two public datasets demonstrate the superiority of our model for distantly supervised relation extraction.

Abstract Auto-regressive extractive summarization approaches determine sentence extraction probability conditioning on previous decisions by maintaining a partial summary representation. Despite its popularity, the framework has two main drawbacks: 1) the partial summary representation is irresolutely denoted by a weighted summation of all the processed sentences without any filtering, resulting in a noisy representation and degrading the effectiveness of extracting subsequent sentences; 2) earlier sentences are biased towards a higher extraction probability due to the sequential nature of sequence tagging. To address these two problems, we propose the Auto-regressive Extractive Summarization with Replacement (AES-Rep), a novel auto-regressive extractive summarization model. In particular, the AES-Rep model consists of two main modules: the extraction decision module that determines whether a sentence should be extracted, and the replacement locater module that enables extracted deficient sentences to be replaced with latter sentences by comparing their expressiveness with respect to the main idea of the document. These modules update the partial summary with explicit actions using elaborated multidimensional guidance. We conduct extensive experiments on the benchmark CNN and DailyMail datasets. Experimental results show that AES-Rep can achieve better performance compared with various strong baselines in terms of multiple ROUGE metrics.

Entity alignment is to find identical entities in different knowledge graphs. Although embedding-based entity alignment has recently achieved remarkable progress, training data insufficiency remains a critical challenge. Conventional semi-supervised methods also suffer from the incorrect entity alignment in newly proposed training data. To resolve these issues, we design an iterative cycle-teaching framework for semi-supervised entity alignment. The key idea is to train multiple entity alignment models (called aligners) simultaneously and let each aligner iteratively teach its successor the proposed new entity alignment. We propose a diversity-aware alignment selection method to choose reliable entity alignment for each aligner. We also design a conflict resolution mechanism to resolve the alignment conflict when combining the new alignment of an aligner and that from its teacher. Besides, considering the influence of cycle-teaching order, we elaborately design a strategy to arrange the optimal order that can maximize the overall performance of multiple aligners. The cycle-teaching process can break the limitations of each model's learning capability and reduce the noise in new training data, leading to improved performance. Extensive experiments on benchmark datasets demonstrate the effectiveness of the proposed cycle-teaching framework, which significantly outperforms the state-of-the-art models when the training data is insufficient and the new entity alignment has much noise.

Neural Relation Extraction (RE) models need large amounts of labeled data for effective training, which mainly comes from automatically labeling by Distant Supervision (DS). Though fast and easy, the label shift problem inevitably happens, i.e., the label distribution of DS-generated training set is quite different from that of the real world (i.e. test set). According to our observations, label shift not only leads to performance diminishment, but also hinders the reliability of DS-RE models by causing bad confidence estimation. In this paper, we make contributions by answering the following three questions: 1) How to improve performance of DS-RE models under label shift? 2) How to make sure their reliability under label shift? 3) How to improve both performance and reliability for DS-RE models under label shift? To the best of our knowledge, this is the first paper to study the performance as well as reliability of DS-RE models under label shift. Experiment results show significant improvements on two real-world datasets and six popular neural RE models, making a step further towards high-performance and reliable RE system under real-world label-shift conditions.

Entity alignment is a crucial task in knowledge graph fusion. However, most entity alignment approaches have the scalability problem. Recent methods address this issue by dividing large KGs into small blocks for embedding and alignment learning in each. However, such a partitioning and learning process results in an excessive loss of structure and alignment. Therefore, in this work, we propose a scalable GNN-based entity alignment approach to reduce the structure and alignment loss from three perspectives. First, we propose a centrality-based subgraph generation algorithm to recall some landmark entities serving as the bridges between different subgraphs. Second, we introduce self-supervised entity reconstruction to recover entity representations from incomplete neighborhood subgraphs, and design cross-subgraph negative sampling to incorporate entities from other subgraphs in alignment learning. Third, during the inference process, we merge the embeddings of subgraphs to make a single space for alignment search. Experimental results on the benchmark OpenEA dataset and the proposed large DBpedia1M dataset verify the effectiveness of our approach.

Document-level Relation Extraction (RE) is a promising task aiming at identifying relations of multiple entity pairs in a document. However, in most cases, a relational fact can be expressed enough via a small subset of sentences from the document, namely evidence sentence. Moreover, there often exist strong semantic correlations between evidence sentences that collaborate together to describe a specific relation. To address these challenges, we propose a novel evidence-aware model for document-level RE. Particularly, we formulate evidence sentence selection as a sequential decision problem through a crafted reinforcement learning mechanism. Considering the explosive search space of our agent, an efficient path searching strategy is executed on the converted document graph to heuristically obtain hopeful sentences and feed them to reinforcement learning. Finally, each entity pair owns a customized-filtered document for further inferring the relation between them. We conduct various experiments on two document-level RE benchmarks and achieve a remarkable improvement over previous competitive baselines, verifying the effectiveness of our method.

We present CLIDSUM, a benchmark dataset towards building cross-lingual summarization systems on dialogue documents.It consists of 67k+ dialogue documents and 112k+ annotated summaries in different target languages.Based on the proposed CLIDSUM, we introduce two benchmark settings for supervised and semi-supervised scenarios, respectively.We then build various baseline systems in different paradigms (pipeline and end-to-end) and conduct extensive experiments on CLIDSUM to provide deeper analyses.Furthermore, we propose mDIALBART which extends mBART via further pre-training, where the multiple objectives help the pre-trained model capture the structural characteristics as well as key content in dialogues and the transformation from source to the target language.Experimental results show the superiority of mDIALBART, as an end-to-end model, outperforms strong pipeline models on CLIDSUM.Finally, we discuss specific challenges that current approaches faced with this task and give multiple promising directions for future research.

Recently, contrastive learning alleviates data sparsity issues and improves the performance of the Graph Neural Network (GNN) recommender models by employing graph structure dropout augmentations. However, these models still face following limitations: (1) Information loss. Dropout may discard helpful information. (2) Insufficient utilization of path-level information. Meta-path is carried numerous high-order information, which has not been well considered in these models. To this end, in this paper, we propose a novel framework, Node and Meta-Path Contrastive Learning for Recommender Systems (NPCRS), which utilizes meta-path to capture path-level information for model learning. Specifically, our approach first generates a meta-path view on the user-item bipartite graph by leveraging meta-path instead of random dropout. Then, we learn the node representation on a user-item bipartite graph and meta-path view to capture both node and path-level information simultaneously. Further, a multi-positive sample mechanism is introduced to define positive and negative samples for contrastive learning. Finally, NPCRS utilizes contrastive learning to learn a more informative node representation. We evaluate the proposed model using three real-world datasets and our experimental results show that our model significantly outperforms the state-of-the-art approaches.

Increased needs for social security promote the development of video surveillance, appealing to the exigency of real-time detection of anomalous events. Considering the rarity and unpredictability of anomalous events, a classical strategy is to model normal data and detect outliers to the model. As a fundamental generative model for time series data, Hidden Markov models (HMM) have been employed in various fields such as speech recognition and video analysis. In this paper, we propose the use of Bayesian HMMs with Dirichlet mixtures which are arrayed along patched frames with Dirichlet distributions as emission probability functions. These spatially-aligned HMMs evolve in parallel, significantly reducing inference time. Learning algorithm based on Stochastic Variational Inference and Discrete Variable Enumeration is applied to our model for fast and robust inference. Experiments over the public UCSD dataset demonstrate the validity of this approach.

Due to fixed convolution kernel structure, it is intractable for Convolutional Neural Networks (CNN) to adequately deal with intact instance geometric transformations, which may paralyze the CNN’s paradigmatic feature extraction. Inspired by Deformable Convolutional Networks, in this work, we introduce stochastic deformatble kernels, where kernel shape is assumed to be random vaiable. Stochastic deformable kernels may extract non-local alienated structural features, which slightly condones geometric transformations. Variational learning of the model based on two approximate posteriors is derived. And experiments over transformed MNIST dataset demonstrate the validity of this approach.

Sports game summarization aims at generating sports news from live commentaries. However, existing datasets are all constructed through automated collection and cleaning processes, resulting in a lot of noise. Besides, current works neglect the knowledge gap between live commentaries and sports news, which limits the performance of sports game summarization. In this paper, we introduce K-SportsSum, a new dataset with two characteristics: (1) K-SportsSum collects a large amount of data from massive games. It has 7,854 commentary-news pairs. To improve the quality, K-SportsSum employs a manual cleaning process; (2) Different from existing datasets, to narrow the knowledge gap, K-SportsSum further provides a large-scale knowledge corpus that contains the information of 523 sports teams and 14,724 sports players. Additionally, we also introduce a knowledge-enhanced summarizer that utilizes both live commentaries and the knowledge to generate sports news. Extensive experiments on K-SportsSum and SportsSum datasets show that our model achieves new state-of-the-art performances. Qualitative analysis and human study further verify that our model generates more informative sports news.

Abstract Cross-lingual summarization is the task of generating a summary in one language (e.g., English) for the given document(s) in a different language (e.g., Chinese). Under the globalization background, this task has attracted increasing attention of the computational linguistics community. Nevertheless, there still remains a lack of comprehensive review for this task. Therefore, we present the first systematic critical review on the datasets, approaches, and challenges in this field. Specifically, we carefully organize existing datasets and approaches according to different construction methods and solution paradigms, respectively. For each type of dataset or approach, we thoroughly introduce and summarize previous efforts and further compare them with each other to provide deeper analyses. In the end, we also discuss promising directions and offer our thoughts to facilitate future research. This survey is for both beginners and experts in cross-lingual summarization, and we hope it will serve as a starting point as well as a source of new ideas for researchers and engineers interested in this area.

Sports game summarization aims to generate news articles from live text commentaries. A recent state-of-the-art work, SportsSum, not only constructs a large benchmark dataset, but also proposes a two-step framework. Despite its great contributions, the work has three main drawbacks: 1) the noise existed in SportsSum dataset degrades the summarization performance; 2) the neglect of lexical overlap between news and commentaries results in low-quality pseudo-labeling algorithm; 3) the usage of directly concatenating rewritten sentences to form news limits its practicability. In this paper, we publish a new benchmark dataset SportsSum2.0, together with a modified summarization framework. In particular, to obtain a clean dataset, we employ crowd workers to manually clean the original dataset. Moreover, the degree of lexical overlap is incorporated into the generation of pseudo labels. Further, we introduce a reranker-enhanced summarizer to take into account the fluency and expressiveness of the summarized news. Extensive experiments show that our model outperforms the state-of-the-art baseline.

The development of existing extractive summarization models for long-form document summarization is hindered by two factors: 1) the computation of the summarization model will dramatically increase due to the sheer size of the input long document; 2) the discourse structural information in the long-form document has not been fully exploited. To address the two deficiencies, we propose HEROES, a novel extractive summarization model for summarizing long-form documents with rich discourse structural information. In particular, the HEROES model consists of two modules: 1) a content ranking module that ranks and selects salient sections and sentences to compose a short digest that empowers complex summarization models and serves as its input; 2) an extractive summarization module based on a heterogeneous graph with nodes from different discourse levels and elaborately designed edge connections to reflect the discourse hierarchy of the document and restrain the semantic drifts across section boundaries. Experimental results on benchmark datasets show that HEROES can achieve significantly better performance compared with various strong baselines.

Zero pronoun resolution aims at recognizing dropped pronouns and pointing out their anaphoric mentions, while non-zero coreference resolution targets at clustering mentions referring to the same entity.Existing efforts often deal with the two problems separately regardless of their close essential correlations.In this paper, we investigate the possibility of jointly solving zero pronoun resolution and coreference resolution via a novel end-to-end neural model.Specifically, we design a gapmasked self-attention model that encodes gaps and tokens in the same space, where gaps could capture valuable contextual information according to their surrounding tokens while tokens could maintain original sequential information without disturbance.Additionally, we also propose a two-stage interaction mechanism to make full use of the exclusive relationship between zero pronouns and mentions.Our empirical study conducted on the OntoNotes 5.0 Chinese dataset shows that our model could outperform corresponding stateof-the-art approaches on both tasks.

Distant supervision is an efficient way to generate large-scale training data for relation extraction without human efforts. However, a coin has two sides. The automatically annotated labels for training data are problematic, which can be summarized as multi-instance multi-label problem and coarse-grained (bag-level) supervised signal. To address these problems, we propose two reasonable assumptions and craft reinforcement learning to capture the expressive sentence for each relation mentioned in a bag. More specifically, we extend the original expressed-at-least-once assumption to multi-label level, and introduce a novel express-at-most-one assumption. Besides, we design a fine-grained reward function, and model the sentence selection process as an auction where different relations for a bag need to compete together to achieve the possession of a specific sentence based on its expressiveness. In this way, our model can be dynamically self-adapted, and eventually implements the accurate one-to-one mapping from a relation label to its chosen expressive sentence, which serves as training instances for the extractor. The experimental results on a public dataset demonstrate that our model constantly and substantially outperforms current state-of-the-art methods for relation extraction.

Currently, most advanced object detectors equip with deep convolutional neural network (CNN) backbones(e.g., ResNet-101) for capturing strong feature representation, but which leads to suffer heavy computational burden. Inversely, some detectors based on lightweight backbones implement real-time processing while accuracy is often not satisfactory. In this paper, we explore an approach to construct a light yet powerful detector by using efficient lightweight backbone (e.g., MobileNet) with our proposed Feature Fusion Block (FFB), composed of Feature Aggregation Block (FAB) and Dense Feature Pyramid (DFP). The extensive experiments confirmed that our proposed Feature Fusion Block Network (FFBNet) is able to improve the accuracy significantly of MobileNet-SSD as well as maintain a close processing speed. Specifically, on Pascal VOC datasets, FFBNet achieves 73.54 mAP at speed of 185 FPS and surpasses MobileNet-SSD five points. Moreover, we apply VGG16 as backbone to further indicate the effectiveness of FFB, which reaches 80.2 mAP on Pascal VOC benchmark. Code is available at https://github.com/fanbinqi/FFBNet.

The recent art in relation extraction is distant supervision which generates training data by heuristically aligning a knowledge base with free texts and thus avoids human labelling. However, the concerned relation mentions often use the bag-of-words representation, which ignores inner correlations between features located in different dimensions and makes relation extraction less effective. To capture the complex characteristics of relation expression and tighten the correlated features, we attempt to discover and utilise informative correlations between features by the following four phases: 1) formulating semantic similarities between lexical features using the embedding method; 2) constructing generative relation for lexical features with different sizes of side windows; 3) computing correlation scores between syntactic features through a kernel-based method; and 4) conducting a distillation process for the obtained correlated feature pairs and integrating informative pairs with existing relation extraction models. The extensive experiments demonstrate that our method can effectively discover correlation information and improve the performance of state-of-the-art relation extraction methods.