With the development of generative models, dialogue systems play an important role in many web applications, such as E-commerce and Question-Answering websites. The accurate user satisfaction estimation (USE) is a critical problem in measuring the quality of dialogue systems. In e-commerce, users usually seek consultation through dialogue systems to know detailed information about the products they intend to purchase. Existing studies mainly focus on analyzing user sentiment in a dialogue for USE, neglecting to understand what the user is concerned about when requesting a consultation. It may cause fatal errors when the response is emotionally friendly but non-informative. Thus, to evaluate how a dialogue satisfies the user's requirements, it is essential to have a conceptual understanding of the products to determine if the response has addressed the user's question. In this paper, we propose a knowledge-enhanced USE model named CoRe-USE, which introduces the Conceptual Knowledge Reasoning for USE in E-Commerce Dialogue Systems. We first design a simple yet efficient entity linking and relation selection module enabling conceptual reasoning in each dialogue. Then, we propose a hierarchical encoder to capture the contextual information in multi-turn dialogues. Finally, we introduce a knowledge enhancement module to fuse conceptual reasoning into contextual embeddings to produce USE. For evaluation, we conduct experiments on three real-world datasets in various scenarios, the results demonstrate the effectiveness and robustness of CoRe-USE compared with SOTA baselines.

In logistic systems, demand prediction is an essential task providing the basis for improving the quality of terminal services, such as pick-up and delivery efficiency. However, the geographical scope of operations across multiple cities brings challenges due to the sparsity of user behavior data, hindering accurate predictions. Despite cross-city prediction methods potentially solving this problem by relying on the label of overlapping users in different cities, annotating these overlapping users is expensive. Additionally, the dynamic and diverse nature of user behaviors complicates feature transfer between cities. In this work, we define the logistics demand prediction problem as forecasting pick-up and delivery demand for zones, the smallest operational units in logistics systems, in different cities. To address the challenge, we propose TSTL, a Transferable Spatio-Temporal Learning framework for cross-city logistics prediction with sparse user data. TSTL advances existing methods from two aspects: (1) User-level invariant representation module extracts consistent user representations for overlapping and non-overlapping users across cities. (2) User-zone graph aggregation module enhances user embeddings by integrating dynamic interactions, such as logistics behaviors, into inherent user relations. Finally, the multi-city transfer module fine-tunes model parameters for city-invariant knowledge adoption and predicts future logistics demand. We implement and evaluate TSTL on one of the largest logistics systems. Extensive offline experiments and real-world deployment demonstrate the effectiveness of TSTL.

Spatio-temporal prediction focuses on jointly modeling spatial correlations and temporal evolution and has a wide range of applications. Due to the heterogeneity of spatio-temporal data, accurate prediction relies on effectively integrating topological structures and sequential patterns. Although recurrent graph learning methods excel at capturing dynamic graph patterns, explicitly inferring future snapshots from historical dynamic graphs remains a significant challenge. Recently, prompt-based graph learning has shown the potential to improve future snapshot inference by leveraging node or task-specific prompts. However, these methods fail to fully capture edge information resulting in incomplete and less accurate representations of future snapshot structures. To bridge this gap, we propose ProST, a framework that Prompts future snapshots on dynamic graphs for Spatio-Temporal prediction, which leverages dynamic graph pre-training to generate a premise graph containing historical graph information and then employs prompts on the premise graph to infer explicit future snapshots. Specifically, this framework comprises three steps: Firstly, dynamic graph pre-training is performed using multi-granularity evolution graph convolution to obtain the premise graph with both local and global features of dynamic graphs. Secondly, prompt subgraphs are used to prompt node pairs and edge features within the premise graph. The subgraph prompt aggregation mechanism propagates this information to generate future snapshots. Finally, we freeze the parameters of the pre-trained model and update the subgraph prompt parameters using meta-learning to adapt to downstream spatio-temporal prediction tasks. Extensive experiments on real-world datasets validate that ProST achieves state-of-the-art performance.

Advancements in Multimodal Large Language Models (MLLMs) have significantly improved information extraction and retrieval performance. Despite these achievements, MLLMs still suffer from the visual object hallucination problem, where models produce plausible, yet incorrect, or irrelevant content not present in the input data. This issue arises from an over-reliance on ''bag-of-objects'' representations and language priors, leading to inadequate extraction of visual objects, along with their attributes and relationships. Existing methods to mitigate these hallucinations are limited by the significant human labor required and the coarse-grained nature. To overcome these challenges, we introduce Multimodal Contrastive Decoding (MMCD), a novel decoding approach that integrates graph-structured reasoning paths with contrastive decoding. MMCD mitigates object hallucinations induced by language priors and enhances the ability of MLLMs to extract and understand compositional information, without additional training or the usage of external tools. This is achieved by masking key objects in images, constructing perturbed scene graphs of attributes and relationships, then contrasting these with the original image and scene graph. Extensive evaluation across three distinct multimodal compositional reasoning tasks: spatial relationship reasoning, alignment of synthetic image and caption, and fine-grained object attribute understanding, show that MMCD consistently surpasses existing decoding methods when applied to various MLLMs. Moreover, MMCD achieves state-of-the-art performance on multiple benchmarks, including the What's Up, SeeTrue and SugarCrepe datasets.

A warehouse-distribution integration (WDI) e-commerce platform is an approach that combines warehousing and distribution processes, which is increasingly adopted in industry to enhance business efficiency. In the WDI e-commerce, one of the most important problems is to estimate the full-link delivery time for decision-making. Traditional methods designed for separate warehouse-distribution models struggle to address challenges in integrated systems. The difficulties stem from two main factors: (i) the contextual influence exerted by neighboring units within heterogeneous delivery networks, and (ii) the uncertainty in delivery times caused by dynamic and periodic temporal factors such as fluctuations in online sales volumes and the varying characteristics of different delivery units (e.g., warehouses and sorting centers). To address these challenges, we propose a novel full-link delivery time estimation framework calledHeterogeneousPeriodicSpatial-TemporalGraphTransformer (HPST-GT). First, we develop heterogeneous graph transformers to capture the hierarchical and diverse information of the warehouse-distribution network. Next, we design spatial-temporal transformers based on heterogeneous features to analyze the correlation between spatial and temporal information. Finally, we create a heterogeneous spatial-temporal graph prediction module to estimate full-link delivery time. Our method, evaluated on a one-month dataset from a leading e-commerce platform, surpasses current benchmarks across multiple performance metrics.

Human behavior prediction is an essential AI-based task, which has inspired many real-world applications. In last-mile logistics, predicting couriers' behavior can benefit the couriers' preference learning and workflow optimization. In this paper, we devote to the behavioral prediction of courier workload and quantify their workload by the working time spent at each area of interest (AOI). Considering the behavior interpretability of inverse reinforcement learning (IRL), existing studies have applied IRL to some real-world transportation prediction scenarios. However, in last-mile logistics, the platform assigns multiple orders to each courier, and couriers also receive new tasks in real-time, which additionally influence the couriers' subsequent decisions. The uncertainty in decision spaces and dynamic the workflow distribution make it more challenging to predict the couriers' working time. In this paper, we propose CourIRL, a practical IRL-based framework leveraging cross-attention to integrate Couriers' historical and spatio-temporal features to predict their future working time. CourIRL formulates the couriers' pick-up and delivery tour as a sequential decision-making process and designs a model-free IRL to learn decision-making preference vectors. A multi-head cross-attention mechanism-based deep regression model is proposed for fine-grained working-time prediction. The results of extensive experiments on two real-world datasets demonstrate that the proposed CourIRL surpasses the state-of-the-art baselines by an average of 6.11% across settings, showing the efficacy and potential contributions of CourIRL in last-mile logistics.

With the development of the logistics industry, the user base of logistics services has expanded swiftly. This rapid increase in user scale presents significant challenges for logistics business management. A fundamental issue in such scenarios is audience expansion, which aims to find users willing to sign long-term services with logistics companies to foster business growth. Existing methods in addressing audience expansion mainly assume user modeling is entangled and neglects the inherent community structure among users. Due to these limitations, the effectiveness of traditional methods in achieving accurate user expansion is often restricted. Our work introduces a novel heterogeneous graph-based model, named Hi-DGN, which concentrates on the Hierarchical information propagation and aggregation in Disentangled Graph Networks for audience expansion. It consists of three main components: (i) the disentangled embedding layer to decouple user representations into different aspects, enabling the extraction of differentiated features; (ii) the hierarchical information propagation module partitions individual nodes into distinct groups and propagates information from group nodes to individual nodes hierarchically to capture diverse granularity representations; and (iii) the aggregation module to fuse all relation-specific embeddings to generate global node embeddings. Extensive experiments on two real-world datasets demonstrate the effectiveness of our method in various evaluation settings.

Recommendation systems play a crucial role in both industrial applications and research fields, which target to understand user preferences and intentions to provide personalized services. Compared to conventional recommendations, repurchase recommendations aim to suggest suitable products to users that they used to buy based on their intention evolution. Existing research on product recommendation can mainly be divided into behavior sequence-based methods and graph-based methods. Although these methods represent user interests and preference features effectively, they still fail to model repurchase behaviors because (i) the environment causing repurchase intention change is neglected and (ii) the lack of feedback after purchasing makes it difficult to learn the impacts of diverse behaviors. To comprehensively consider these limitations, we design a D ual I ntention-aware F usion N etwork framework (DIFN) to understand the effects of environment and after-purchasing feedback on users' intentions. Firstly, a hierarchical graph-based multi-level relational attention module is designed to effectively extract basic user features and spatial features from complex environmental information. Then, we introduce a behavior intention module and a usage intention module for different types of feedback data. Finally, we propose a dual intention fusion network that effectively fuses user basic features with spatial attributes and user intention features with temporal attributes for recommendation. Comprehensive evaluations on real-world datasets show that our method exceeds state-of-the-art baselines, which show an average of 8.2% improvements in different metrics.

Integrated Warehousing and Distribution Supply Networks (IWDSN) have shown their high efficiency in E-commerce. Efficient supply capacity prediction is crucial for logistics systems to maintain the delivery capacity to meet users' requirements. However, unforeseen events such as extreme weather and public health emergencies pose challenges in supply forecasting. Previous work mainly infers supply optimization based on the invariant topology of logistic networks, neglecting dynamic routing and distinct node effects reacting to emergencies. To address these challenges, the hierarchical relations among warehouses, sorting centers, and delivery stations in logistic networks are necessary to learn the diverse reactions. In this paper, we propose a hierarchical spatio-temporal graph learning model to predict the emergency supply capacity of IWDSN based on micro and macro graphs. The micro graph shows transportation connectivity while the macro graph shows the geographical correlation. Specifically, it consists of three components. (1) For micro graphs, a metapath aggregation strategy is designed to capture dynamic routing information on both route-view and event-view graphs. (2) For macro graphs, a bipartite graph learning approach to extract spatial representations. (3) For spatio-temporal feature fusion, the spatio-temporal joint forecasting module combines the temporal feature from the time-series encoder with hierarchical spatial features to predict the future supply capacity. The extensive experiments on two real-world datasets demonstrate the effectiveness of our proposed model, which achieves state-of-the-art performance compared with advanced baselines.

Trajectory data is a valuable asset for service management and spatio-temporal mining in transportation and logistics systems. However, due to equipment failure, network delay, and energy constraints, some trajectory point may be missed, which makes it difficult for trajectory-based management. Some researchers have focused on recovering sparse trajectories from road networks and historical trajectory data, but these methods are ineffective when the road network is incomplete. Recent research works have explored learning-based methods to recover trajectories in free space but lack user movement behavior modeling and efficient feature extraction on sparse long-range trajectories. Our work exploits the periodic behavior of couriers and fine-grained Area of Interest (AOI) data for sparse trajectory recovery in last-mile delivery. However, we face challenges with AOI access sequence deviations due to GPS inaccuracies and abnormal courier behaviors, as well as the complex, dynamic relationships within and between courier routes due to uncertain pick-up demands. To address these challenges, we design a graph-based multi-task learning framework, focusing on multi-scale attention fusion for end-to-end free space trajectory recovery. Our approach starts with a behavior-aware graph network that generates detailed spatial features. Following this, we propose a multi-scale attention fusion mechanism to extract intra- and inter-trajectory features. Finally, we design a multi-task learning module that predicts both coarse-grained spatial access sequences and fine-grained trajectory points. We evaluate the model with six-month data involved with more than 360,000 trajectory segments and more than 7.2 million waybills collected from one of the largest logistic companies in China. Extensive experiments on real-world datasets demonstrate that our method outperforms state-of-the-arts in multiple metrics.

Recently, integrated warehouse and distribution logistics systems are widely used in E-commerce industries to adjust to constantly changing customer demands. It makes the prediction of purchase demand and delivery supply capacity a crucial problem to streamline operations and improve efficiency. The interaction between such demand and supply not only relies on their economic relationships but also on consumer psychology caused by daily events, such as epidemics, promotions, and festivals. Although existing studies have made great efforts in the joint prediction of demand and supply considering modeling the demand-supply interactions, they seldom refer to the impacts of diverse events. In this work, we propose MulSTE, a Multi-view Spatio-Temporal learning framework with heterogeneous Event fusion. Firstly, an Event Fusion Representation (EFR) module is designed to fuse the textual, numerical, and categorical heterogeneous information for emergent and periodic events. Secondly, a Multi-graph Adaptive Convolution Recurrent Network (MGACRN) is developed as the spatio-temporal encoder (ST-Encoder) to capture the evolutional features of demand, supply, and events. Thirdly, the Event Gated Demand-Supply Interaction Attention (EGIA) module is designed to model the demand-supply interactions during events. The evaluations are conducted on two real-world datasets collected from JD Logistics and public websites. The experimental results show that our method outperforms state-of-the-art baselines in various metrics.

Multimodal content generation, which leverages visual information to enhance the comprehension of cross-modal understanding, plays a critical role in Multimodal Information Retrieval. With the development of large language models (LLMs), recent research has adopted visual instruction tuning to inject the knowledge of LLMs into downstream multimodal tasks. The high complexity and great demand for resources urge researchers to study efficient distillation solutions to transfer the knowledge from pre-trained multimodal models.(teachers) to more compact student models. However, the instruction tuning for knowledge distillation in multimodal LLMs is resource-intensive and capability-restricted. The comprehension of students is highly reliant on the teacher models. To address this issue, we propose a novel Multimodal Distillation Calibration framework (MmDC). The main idea is to generate high-quality training instances that challenge student models to comprehend and prompt the teacher to calibrate the knowledge transferred to students, ultimately cultivating a better student model in downstream tasks. This framework comprises two stages: (1) multimodal alignment and (2) knowledge distillation calibration. In the first stage, parameter-efficient fine-tuning is used to enhance feature alignment between different modalities. In the second stage, we develop a calibration strategy to assess the student model's capability and generate high-quality instances to calibrate knowledge distillation from teacher to student. The experiments on diverse datasets show that our framework efficiently improves the student model's capabilities. Our 7B-size student model, after three iterations of distillation calibration, outperforms the current state-of-the-art LLaVA-13B model on the ScienceQA and LLaVA Test datasets and also exceeds other strong baselines in a zero-shot setting.

Accurate customer LifeTime Value (LTV) predictions are crucial for customer relationship management, especially in Supply Chain Platforms (SCP), which involve effectively managing the service resources in business decision-making. Previous LTV prediction methods usually rely on ample historical customer data, which is not available in the early stages of a customer's lifecycle. It makes the modeling of the historical customer data a difficult task due to the data sparsity. Besides, the long-tail distribution of customer LTV also brings new challenges to the prediction of LTV. To tackle the above issues, we propose CDLtvS, a novel Cross Domain method for customer Lifetime value prediction in SCP. It leverages rich cross-domain information from upstream platforms to enhance LTV predictions in downstream platforms. Firstly, CDLtvS pre-trains the customer representations by an LTV modeling framework named LtvS in source and target domains separately. Specifically, LtvS incorporates the Expert Mask Network (ExMN), which not only effectively models the long-tail distribution of LTV in single-domain but also resolves cross-domain learning model bias resulting from this distribution. Then, the various-level alignment mechanism is introduced to keep the consistency of knowledge transferring from source to target domains on both sparse and non-sparse data. Comprehensive experiments on real-world data from JD, one of the world's largest supply chain platforms, demonstrate that CDLtvS achieves a normalized mean average error of 0.3378 in LTV prediction, outperforming 16.3% to the baseline. Additionally, the improvements of ≥2.3% across various data sparsity levels (0% -- 80%) provide valuable insights into cross-domain LTV modeling.

In recent years, emergency last-mile logistics (ELML) have played an essential role in urban emergencies. The efficient allocation of couriers in ELML is of practical significance to ensure the supply of essential materials, especially in public health emergencies (PHEs). However, couriers allocation becomes challenging due to the instability of demand, dynamic supply comprehension, and the evolutional delivery environment for ELML caused by PHEs. While existing work has delved into couriers allocation, the impact of PHEs on demand-supply-delivery has yet to be considered. In this work, we design PTOCA, a Predict-Then-Optimize Couriers Allocation framework. Specifically, in the prediction stage, we design a resource-aware prediction module that performs spatio-temporal modeling of unstable demand characteristics using a variational graph GRU encoder and builds a task-resource regressor to predict demand accurately. In the optimization stage, firstly, the priority ranking module solves the matching of delivery resources under demand-supply imbalance. Then the multi-factor task allocation module is used to model the dynamic evolutional environment and reasonably assign the delivery tasks of couriers. We evaluate PTOCA using real-world data covering 170 delivery zones, more than 10,000 couriers, and 100 million delivery tasks. The data is collected from JD Logistics, one of the largest logistics service companies. Extensive experimental results show that our method outperforms the baseline in task delivery rate and on-time delivery rate.

The generalizability to new databases is of vital importance to Text-to-SQL systems which aim to parse human utterances into SQL statements. Existing works achieve this goal by leveraging the exact matching method to identify the lexical matching between the question words and the schema items. However, these methods fail in other challenging scenarios, such as the synonym substitution in which the surface form differs between the corresponding question words and schema items. In this paper, we propose a framework named ISESL-SQL to iteratively build a semantic enhanced schema-linking graph between question tokens and database schemas. First, we extract a schema linking graph from PLMs through a probing procedure in an unsupervised manner. Then the schema linking graph is further optimized during the training process through a deep graph learning method. Meanwhile, we also design an auxiliary task called graph regularization to improve the schema information mentioned in the schema-linking graph. Extensive experiments on three benchmarks demonstrate that ISESL-SQL could consistently outperform the baselines and further investigations show its generalizability and robustness.

Prediction over event sequences is critical for many real-world applications in Information Retrieval and Natural Language Processing. Future Event Generation (FEG) is a challenging task in event sequence prediction because it requires not only fluent text generation but also commonsense reasoning to maintain the logical coherence of the entire event story. In this paper, we propose a novel explainable FEG framework, Coep. It highlights and integrates two types of event knowledge, sequential knowledge of direct event-event relations and inferential knowledge that reflects the intermediate character psychology between events, such as intents, causes, reactions, which intrinsically pushes the story forward. To alleviate the knowledge forgetting issue, we design two modules, IM and GM, for each type of knowledge, which are combined via prompt tuning. First, IM focuses on understanding inferential knowledge to generate commonsense explanations and provide a soft prompt vector for GM. We also design a contrastive discriminator for better generalization ability. Second, GM generates future events by modeling direct sequential knowledge with the guidance of IM. Automatic and human evaluation demonstrate that our approach can generate more coherent, specific, and logical future events.

Text classification in low-resource languages (eg Thai) is of great practical value for some information retrieval applications (eg sentiment-analysis-based restaurant recommendation). Due to lacking large-scale corpus for learning comprehensive text representation, bilingual text classification which borrows the linguistics knowledge from a rich-resource language becomes a promising solution. Despite the success of bilingual methods, they largely ignore another source of semantic information---the writing system. Noting that most low-resource languages are phonographic languages, we argue that a logographic language (eg Chinese) can provide helpful information for improving some phonographic languages' text classification, since a logographic character (ie logogram) could represent a sememe or a whole concept, not only a phoneme or a sound. In this paper, by using a phonographic labeled corpus and its machine-translated logographic corpus both, we devise a framework to explore the central theme of utilizing logograms as a "semantic detection assistant''. Specifically, from a logographic labeled corpus, we first devise a statistical-significance-based module to pick out informative text pieces. To represent them and further reduce the effects of translation errors, our approach is equipped with Gaussian embedding whose covariances serve as reliable signals of translation errors. For a test document, all seeds' Gaussian representations are used to convolute the document and produce a logographic embedding, before being fused with its phonographic embedding for final prediction. Extensive experiments validate the effectiveness of our approach and further investigations show its generalizability and robustness.

Event sequence prediction has wide applications on economics, electronic health and social media monitoring. Accurate prediction of event sequences can help provide better service to customers and prevent risks. Recent works try to address the problem aiming at learning the impact of past events on the future events using deep learning methods. Such works often take the past event sequences as input and model the self-change transformations of the events, and few of them concerned the effect of event attributes. We propose an RNN-based predictive model to encode multiple attributes as attached information of the event for predicting next event and its attributes given past sequences. To learn how important each attribute is for the event, we design a component modulator to customize weights for representations of the event and its attributes. The more important the information is, the relevant weight will be higher. Finally, the prediction of next event and its attributes are conducted simultaneously with a different modulator for each predictive task. The performance of the proposed model was evaluated on 5 real-life datasets, containing two different types of event logs. The results show that our model outperforms the baselines and the state-of-the-art, not only on the prediction of next event and its attributes but also the generation of event sequence suffix.