Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Xiaoming LiSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works205 from public data
- InfoXLM: An Information-Theoretic Framework for Cross-Lingual Language Model Pre-Training454
An information-theoretic framework that formulates cross-lingual language model pre- training as maximizing mutual information between multilingual-multi-granularity texts is presented and a new pre-training task based on contrastive learning is proposed.
- Multi-level Cross-view Contrastive Learning for Knowledge-aware Recommender System259
This paper proposes a novel multi-level cross-view contrastive learning mechanism, named MCCLK, which comprehensively considers three different graph views for KG-aware recommendation, including global-level structural view, local-level collaborative and semantic views, and a k-Nearest-Neighbor item-item semantic graph construction module is proposed.
- 186
- Cross-Lingual Natural Language Generation via Pre-Training142
Experimental results on question generation and abstractive summarization show that the model outperforms the machine-translation-based pipeline methods for zero-shot cross-lingual generation and improves NLG performance of low-resource languages by leveraging rich-resource language data.
- XLM-E: Cross-lingual Language Model Pre-training via ELECTRA141
This paper introduces ELECTRA-style tasks to cross-lingual language model pre-training, and pretrain the model, named as XLM-E, on both multilingual and parallel corpora, which outperforms the baseline models on various cross-lingsual understanding tasks with much less computation cost.
- Multi-View Intent Disentangle Graph Networks for Bundle Recommendation130
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 is proposed.
- Improving Knowledge-aware Recommendation with Multi-level Interactive Contrastive Learning120
This paper focuses on exploring contrastive learning in KGR and proposes a novel multi-level interactive contrastiveLearning mechanism, which conducts layer-wise self-supervised learning by contrasting layers of different parts within graphs, which is also an "interaction" action.
- Read, Listen, and See: Leveraging Multimodal Information Helps Chinese Spell Checking118
The ReaLiSe model tackles the CSC task by capturing the semantic, phonetic and graphic information of the input characters, and selectively mixing the information in these modalities to predict the correct output.
- BiSyn-GAT+: Bi-Syntax Aware Graph Attention Network for Aspect-based Sentiment Analysis106
A Bi-Syntax aware Graph Attention Network that fully exploits the syntax information of the constituent tree of a sentence to model the sentiment-aware context of every single aspect and the sentiment relations across aspects (called inter-context) for learning.
- MLS3RDUH: Deep Unsupervised Hashing via Manifold based Local Semantic Similarity Structure Reconstructing103
This paper proposes a novel deep unsupervised hashing method, called MLS3RDUH, which can reduce the noisy datapoints to further enhance retrieval performance and outperforms the state-of-the-art baselines.
- Deep Cross-Modal Hashing With Hashing Functions and Unified Hash Codes Jointly Learning99
A novel end-to-end Deep Cross-Modal Hashing with Hashing Functions and Unified Hash Codes Jointly Learning (DCHUC) is proposed, which outperforms the state-of-the-art cross-modal hashing methods.
- Improving Pretrained Cross-Lingual Language Models via Self-Labeled Word Alignment87
This paper introduces denoising word alignment as a new cross-lingual pre-training task that can serve as a pretrained word aligner, which achieves reasonably low error rate on the alignment benchmarks.
- Food recommendation with graph convolutional network83
This work proposes a novel model Food recommendation with Graph Convolutional Network (FGCN), which exploits ingredient-ingredient, ingredient-recipe, and recipe-user relations deeply and could alleviate the sparsity issue in food recommendation.
- mT6: Multilingual Pretrained Text-to-Text Transformer with Translation Pairs82
Experimental results show that the proposed mT6 improves cross-lingual transferability over mT5, and proposes a partially non-autoregressive objective for text-to-text pre-training.
- SSHLDA: A Semi-Supervised Hierarchical Topic Model65
A semi-supervised hierarchical topic model which aims to explore new topics automatically in the data space while incorporating the information from observed hierarchical labels into the modeling process, called Semi-Supervised Hierarchical Latent Dirichlet Allocation (SSHLDA).
- Rethinking Task-Oriented Dialogue Systems: From Complex Modularity to Zero-Shot Autonomous Agent62
This paper proposes a novel fully zero-013 shot 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 lan-017 guage model (e.g. GPT-4).
- 61
- 58
- Multi-level Contrastive Learning Framework for Sequential Recommendation57
The proposed MCLSR outperforms the state-of-the-art methods consistently and 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).
- 53
- Automatic labeling hierarchical topics51
This paper proposes two effective algorithms that automatically assign concise labels to each topic in a hierarchy by exploiting sibling and parent-child relations among topics and shows that the inter-topic relation is effective in boosting topic labeling accuracy.
- 48
- Partial-Softmax Loss based Deep Hashing45
A novel Partial-Softmax Loss based Deep Hashing, called PSLDH, is proposed to generate high-quality hash codes and outperforms the state-of-the-art baselines in image retrieval task.
- Person-job fit estimation from candidate profile and related recruitment history with co-attention neural networks41
A novel neural network approach for person- Job fit is proposed, which estimates person-job fit from candidate profile and related recruitment history with co-attention neural networks (named PJFCANN), and the historical successful recruitment records are introduced to enrich the features of resumes and job postings and strengthen the current matching process.
- 40
- 40
- Are Noisy Sentences Useless for Distant Supervised Relation Extraction?39
A novel method for distant supervised relation extraction is proposed, which employs unsupervised deep clustering to generate reliable labels for noisy sentences, which outperforms the state-of-the-art baselines on a popular benchmark dataset, and can indeed alleviate the noisy labeling problem.
- Copy Is All You Need37
This paper formulate text generation as progressively copying text segments from an existing text collection, and compute the contextualized representations of meaningful text segments and index them using efficient vector search toolkits.
- Data-Aware Proxy Hashing for Cross-modal Retrieval37
A novel Data-Aware Proxy Hashing for cross-modal retrieval, called DAPH, which exploits the three types of data-aware proxy hash codes to supervise the training of modality-specific hashing networks and is able to generate discriminate hash codes with the semantic information preserved adequately.
- 37
- Improving Personality Consistency in Conversation by Persona Extending35
A novel retrieval-to-prediction paradigm consisting of two subcomponents, namely, Persona Retrieval Model (PRM) and Posterior-scored Transformer (PS-Transformer), that adopts a persona posterior distribution that further considers the actual personas used in the ground response, maximally mitigating the gap between training and inferring.
- Multi-view Hypergraph Contrastive Policy Learning for Conversational Recommendation33
A novel model, namely Multi-view Hypergraph Contrastive Policy Learning (MHCPL), which timely chooses useful social information according to the interactive history and builds a dynamic hypergraph with three types of multiplex relations from different views is proposed.
- Context-aware Entity Typing in Knowledge Graphs33
A novel method to infer entities' missing types in knowledge graphs by utilizing entities' contextual information is proposed and a novel loss function is proposed to alleviate the false-negative problem during training.
- Position-aware self-attention based neural sequence labeling31
This work proposes an innovative attention-based model as well as a well-designed self-attentional context fusion layer within a neural network architecture, to explore the positional information of an input sequence for capturing the latent relations among tokens.
- PONE30
This article proposes a novel learning-based metric that significantly improves the correlation with human judgments by using augmented POsitive samples and valuable NEgative samples, called PONE, and measures systematically all kinds of metrics to check which kind is best.
- Supervised Deep Hashing for Hierarchical Labeled Data29
A novel deep hashing method, called supervised hierarchical deep hashing (SHDH), is proposed to perform hash code learning for hierarchical labeled data by weighting each level, and a deep neural network is designed to obtain a hash code for each data point.
- Target-guided Emotion-aware Chat Machine28
A unified end-to-end neural architecture is proposed, 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.
- Weighted Gaussian Loss based Hamming Hashing28
A novel Weighted Gaussian Loss based Hamming Hashing, called WGLHH, is proposed, which introduces a weighted Gaussian loss to optimize hashing model and outperforms the state-of-the-art baselines in image retrieval task.
- Joint Multi-Facts Reasoning Network for Complex Temporal Question Answering Over Knowledge Graph27
This paper proposes Joint Multi Facts Reasoning Network (JMFRN), to jointly reasoning multiple temporal facts for accurately answering complex temporal questions, and significantly outperforms the state-of-art on the wellknown complex temporal question benchmark TimeQuestions.
- Exploring global information for session-based recommendation27
This work proposes a basic GNN-based session recommendation model on SBR problem by solely using session-level item-transition information on session graph and proposes a novel approach, called Session-based Recommendation with Global Information (SRGI), which infers the user preferences of the current session via fully exploiting global item- Transition information over all sessions.
- Stack-VS: Stacked Visual-Semantic Attention for Image Caption Generation26
A novel well-designed stack decoder model, which is constituted by a sequence of decoder cells, each of which contains two LSTM-layers work interactively to re-optimize attention weights on both visual-level feature vectors and semantic-level attribute embeddings for generating a fine-grained image caption.
- 23
- 23
- 22
- Conceptualization topic modeling22
The results show that the proposed models significantly outperform the baselines in terms of case study and perplexity, which means the new assumption is more reasonable than traditional one.
- Exploring Dense Retrieval for Dialogue Response Selection21
This work presents a novel and efficient response selection model and a set of tailor-designed learning strategies to train it effectively, which notably outperforms the state-of-the-art baselines on both re-rank and full-rank evaluations.
- Unsupervised Hashing with Semantic Concept Mining20
A novel Unsupervised Hashing with Semantic Concept Mining, called UHSCM, which leverages a VLP model to construct a high-quality similarity matrix, and a novel hashing loss with a modified contrastive loss based regularization item is proposed to optimize the hashing network.
- From Question to Text20
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.
- 19
- 18
- 17
- 16
- Three birds, one stone: A novel translation based framework for joint entity and relation extraction16
- 16
- 16
- S2JSD-LSH: A Locality-Sensitive Hashing Schema for Probability Distributions16
This paper proposes a novel LSH scheme adapted to S2JSD-distance for approximate nearest neighbors search in high-dimensional probability distributions, and defines the specific hashing functions, and proves their local-sensitivity.
- I Know What You Want to Express: Sentence Element Inference by Incorporating External Knowledge Base16
- 15
- Graph neural networks for visual question answering: a systematic review15
This paper systematically reviews various graph neural networks based studies for image-based VQA, and provides some possible suggestions to mitigate still existing challenges for future research in visual question answering.
- 15
- Object Detection based Deep Unsupervised Hashing15
A novel Object Detection based Deep Unsupervised Hashing method is proposed that outperforms the state-of-the-art unsupervised hashing methods in the image retrieval task and is used to guide the learning process to generate high-quality hash codes.
- A Hierarchical Context Augmentation Method to Improve Retrieval-Augmented LLMs on Scientific Papers14
This work proposes a hierarchical context augmentation method, which helps Retrieval-Augmented LLMs to autoregressively learn the structure knowledge of scientific papers, and utilizes the document tree to represent the hierarchical relationship of a paper and enhance the structure information of scientific context.
- Exploiting global contextual information for document-level named entity recognition14
A model called Global Context enhanced Document-level NER (GCDoc) to leverage global contextual information from two levels, i.e., both word and sentence, and employs a cross-sentence module which encodes adjacent sentences and fuses it with the current sentence representation via attention and gating mechanisms.
- 14
- Re-ranking voting-based answers by discarding user behavior biases14
This paper identifies the existence of two kinds of biases; position bias and appearance bias and proposes a joint click model for dealing with both of them and demonstrates how the ranking performance of the proposed model outperforms traditional methods with biases ignored.
- ALOHA: Adapting Local Spatio-Temporal Context to Enhance the Audio-Visual Semantic Segmentation13
The ALOHA framework, which Adapts LOcal spatio-temporal context to enHAnce AVSS and utilizes a parameter-shared encoder and applies the adapters in a staged manner to better adapt to embodied AI systems, significantly reduces the number of trainable parameters, making it more parameter-efficient.
- 13
- Can Monolingual Pretrained Models Help Cross-Lingual Classification?13
Two approaches to improve zero-shot cross-lingual classification, by transferring the knowledge from monolingual pretrained models to multilingual ones are presented.
- 12
- Learning Relation Ties with a Force-Directed Graph in Distant Supervised Relation Extraction12
A novel force-directed graph is proposed to comprehensively learn relation ties and is capable of modeling global correlation and mutual exclusion between relations, and outperforms the state-of-the-art baselines.
- Comprehensive Study: How the Context Information of Different Granularity Affects Dialogue State Tracking?12
This paper studies and discusses how the context information of different granularity affects dialogue state tracking and applies the findings about context granularity to few-shot learning scenario.
- Exploiting Group-Level Behavior Pattern for Session-Based Recommendation11
In RNMSR, a novel Repeat-aware Neural Mechanism for Session-based Recommendation (RNMSR) is proposed, which combines 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.
- 11
- PNE: Label Embedding Enhanced Network Embedding11
A new point of view for semi-supervised NRL is provided and a novel model named Predictive Network Embedding (PNE) is presented, which embed nodes and labels into the same latent space instead of training a classifier in the representation learning process and the discriminability of node representations is enhanced by incorporating the label information.
- High-Performance Signal Detection for Adverse Drug Events using MapReduce Paradigm.11
This study proposed a MapReduce-based algorithm, for common ADE detection approach, Proportional Reporting Ratio (PRR), and tested it in mining spontaneous ADE reports from FDA, and demonstrated that Map Reduce programming model could improve the performance of common signal detection algorithm for pharmacovigilance in a distributed computation environment at approximately liner speedup rates.
- MTAVG-Bench: A Diagnostic Benchmark for Multi-Talker Dialogue-Centric Audio-Video Generation10
MTAVG-Bench is designed to evaluate whether proprietary and open-source omni-models can reliably identify failure modes in multi-speaker T2AV outputs, and benchmark 12 proprietary and open-source omni-models on MTAVG-Bench.
- Hashing based Efficient Inference for Image-Text Matching10
A novel hashing based efficient inference module called HEI is proposed, which can be plugged into the existing framework to speed up the inference step without reducing the retrieval performance.
- 9
- 9
- HSDS: An Abstractive Model for Automatic Survey Generation9
This paper proposes a novel abstractive method named Hierarchical Seq2seq model based on Dual Supervision (HSDS) to solve problems above, given multiple scientific papers in the same research area as input, the model aims to generate a corresponding survey.
- Query completion in community-based Question Answering search9
This work proposes a novel query completion algorithm based on ranking cQA questions with entity and phrase information for cZA search, and a demonstration system has been developed and shows that the proposed algorithm can successfully improve user experience.
- A novel unsupervised method for new word extraction9
A novel method to extract new words that not only considers domain specificity, but also combines with multiple statistical language knowledge to extract the top ranked new words.
- 8
- 8
- 8
- Towards End-to-End Learning for Efficient Dialogue Agent by Modeling Looking-ahead Ability8
This paper proposes 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.
- Labeled Phrase Latent Dirichlet Allocation and its online learning algorithm8
This paper proposes a novel topic model, called Labeled Phrase Latent Dirichlet Allocation (LPLDA), which regards each document as a mixture of phrases and partly considers the word order, and develops a batch inference algorithm based on Gibbs sampling technique for LPLDA.
- SizeSpotSigs: An Effective Deduplicate Algorithm Considering the Size of Page Content8
An algorithm named SizeSpotSigs is presented, taking the size of page contents into account, which could handle both short and long Web pages, and works better than state-of-the-art approaches such as SpotSigs over a demonstrative Mixer of manually assessed nearduplicate news articles.
- 7
- 7
- 7
- Toward Real-life Dialogue State Tracking Involving Negative Feedback Utterances7
Simulation of negative feedback utterances in dialogue state tracking is explored in detail and three aspects of interaction mechanism that should be considered in real-life conversations involving negative Feedback utterances are explored and evaluated.
- Neural Variational Gaussian Mixture Topic Model7
A novel neural topic model called Neural Variational Gaussian Mixture Topic Model (NVGMTM) is proposed, which uses Gaussian distribution to depict the semantic relevance between words in the topics and outperforms the state-of-the-art baselines.
- A Survey of Automatic Evaluation Methods on Text, Visual and Speech Generations6
A comprehensive review and a unified taxonomy of automatic evaluation methods for generated content across all three modalities is presented and five fundamental paradigms that characterize existing evaluation approaches across these domains are identified.
- AgentTOD: A Task-Oriented Dialogue Agent with a Flexible and Adaptive API Calling Paradigm6
This article proposes AgentTOD, a novel TOD framework that uses a large language model (LLM) as the intelligent agent to achieve a flexible dialogue paradigm that has superior performance on TODs as well as a superior adaptability to new task scenarios.
- 6
- Span Graph Transformer for Document-Level Named Entity Recognition6
This work proposes a novel Span Graph Transformer (SGT) method for document-level NER, which constructs long-range contextual dependencies at both the token and span levels and achieves better performance than previous state-of-the-art models.
- 6
- Deep kernel supervised hashing for node classification in structural networks6
A novel Deep Kernel Supervised Hashing (DKSH) method to learn the hashing representations of nodes for node classification is proposed and significantly outperforms the state-of-the-art baselines over three real world benchmark datasets.
- 6
- Locality-Sensitive Hashing for Finding Nearest Neighbors in Probability Distributions6
This paper presents a novel LSH scheme adapted to angular distance for ANN search in high-dimensional probability distributions, and proposes a Sequential Interleaving algorithm based on the “Unbalance Effect” of Euclidean and angular metrics for probability distributions.
- An Empirical Study on the Language Modal in Visual Question Answering5
A series of experiments revealed that apart from prior bias caused by question types, there is a notable influence of postfix-related bias in inducing biases, and training VQA models with word-sequence-related variant questions demonstrated improved performance on the out-of-distribution benchmark.
- Momentum Decoding: Open-ended Text Generation As Graph Exploration5
This study formulate open-ended text generation from a new perspective, i.e., it is viewed as an exploration process within a directed graph, and proposes a novel decoding method -- \textit{momentum decoding} -- which encourages the LM to explore new nodes outside the current graph.
- Unsupervised Question Answering via Answer Diversifying5
Extensive experiments show that the proposed method outperforms previous unsupervised models on five benchmark datasets, including SQuADv1.1, News QA, TriviaQA, BioASQ, and DuoRC, and shows strong performance in the few-shot learning setting.
- Generating Informative Dialogue Responses with Keywords-Guided Networks5
A simple but effective keywords-guided Sequence-to-Sequence model (KW-Seq2Seq) which uses keywords information as guidance to generate open-domain dialogue responses and produces more informative, coherent and fluent responses.
- Labeled Phrase Latent Dirichlet Allocation5
A novel topic model, called Labeled Phrase LDA, which synchronously considers the supervised information and word order is proposed, which significantly outperforms baselines in terms of case study, perplexity and scalability.
- EHLLDA: A Supervised Hierarchical Topic Model5
A supervised hierarchical topic models are introduced: Extended Hierarchical Labeled Latent Dirichlet Allocation (EHLLDA), which aim to relax the assumptions of hLLDA by incorporating prior information of labels into hLLda.
- MTAVG-Bench 2.0: Diagnosing Failure Modes of Cinematic Expressiveness in Multi-Talker Audio-Video Generation4
Results demonstrate that MTAVG-Bench 2.0 provides a systematic benchmark for failure diagnosis in cinematic multi-talker audio-video generation, and establishes a high-level failure taxonomy spanning acting, narrative, atmosphere, and audio-visual language.
- Multi-scale dual-stream visual feature extraction and graph reasoning for visual question answering4
This study proposes a multi-scale dual-stream visual feature extraction method that combines grid and region features to enhance both global and local visual feature representations and proposes a cross-modal self-attention fusion strategy, which enables the model to focus selectively on the most relevant parts of both the image and the question.
- 4
- Beyond Labels and Topics: Discovering Causal Relationships in Neural Topic Modeling4
This paper proposes 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.
- CriticEval: Evaluating Large Language Model as Critic4
C RITIC B ENCH is introduced, a novel benchmark designed to comprehensively and reliably evaluate four key critique ability dimensions of LLMs: feedback, comparison, refinement and meta-feedback.
- A Cross-Modal Classification Dataset on Social Network4
A novel Cross-Modal Classification Dataset constructed from Weibo called CMCD is constructed, which collects tweets with three modalities of text, image and video from 18 general categories, and filters tweets that can easily be classified by only textual contents.
- A Novel Fast Framework for Topic Labeling Based on Similarity-preserved Hashing4
This paper proposes a novel fast topic labeling framework that casts the labeling problem as a k-nearest neighbor (KNN) search problem in a probability vector set, and shows that the proposed sequential interleaving method based on locality sensitive hashing (LSH) technology is efficient in boosting the comparison speed among probability distributions.
- FCL: A New Network Words Extraction Approach Based on Statistical Language Knowledge4
This paper proposes a novel approach of FCL to extract new network words that not only considers domain specificity, but also combines with multiple statistical language knowledge.
- Hierarchical topic integration through semi-supervised hierarchical topic modeling4
This paper proposes to use semi-supervised topic modeling to solve the problem of automatically integrating latent topics in a plain collection with the Topics in a hierarchical structured collection in a principled way.
- A Survey of Web Page Cleaning Research4
All kinds of Web page cleaning techniques and framew o rks, including SST, Shingle, Pagelet, DSE, etc, are summarized, and the experimental datasets and experimental methods used in all kinds of page cleaning methods are described.
- 3
- 3
- Can Cross-Lingual Transferability of Multilingual Transformers Be Activated Without End-Task Data?3
Experimental results show that P LUG I N -X successfully activates the cross-lingual transferability of multilingual Transformers without accessing end-task data.
- 3
- ET5: A Novel End-to-end Framework for Conversational Machine Reading Comprehension3
A novel end-to-end framework for conversational machine reading comprehension based on shared parameter mechanism, called entailment reasoning T5 (ET5), is proposed, which achieves new state-of-the-art results on the ShARC leaderboard with the BLEU-4 score of 55.2.
- Hammer PDF: An Intelligent PDF Reader for Scientific Papers3
The proposed Hammer PDF Reader can help researchers, especially those studying computer science, to improve the efficiency and experience of reading scientific papers.
- Exploiting Repeated Behavior Pattern and Long-term Item dependency for Session-based Recommendation.3
This work proposes a novel Repeat-aware Neural Mechanism for Session-based Recommendation (RNMSR), which contains the potential intent information and critical item frequency signal into SBR, and built a similarity-based session graph based on long-term dependencies within a session.
- 3
- Multi-task Learning for Low-Resource Second Language Acquisition Modeling3
This paper proposes a novel SLA modeling method, which learns the latent common patterns among different language-learning datasets by multi-task learning and are further applied to improving the prediction performance in low-resource scenarios.
- 3
- Supervised Hashing for Multi-labeled Data with Order-Preserving Feature3
This work proposed a novel supervised hashing method to perform hash codes learning for multi-labeled data that extracts the order-preserving data features through deep convolutional neural network and would be used for learning hash codes.
- 3
- An Online Inference Algorithm for Labeled Latent Dirichlet Allocation3
This paper develops an efficient online algorithm for Labeled LDA, called online-LLDA, based on particle filter, a Sequential Monte Carlo approximation technique, which significantly outperforms batch algorithm (batch- LLDA) in time, while preserving equivalent quality.
- 2
- DeepSurvey-Bench: Evaluating Academic Value of Automatically Generated Scientific Surveys2
DeepSurvey-Bench is proposed, a comprehensive benchmark for evaluating the academic value of automatically generated surveys that aligns closely with human assessments in evaluating the academic value of surveys, and reveals underlying academic value beyond the reach of surface-level quality metrics.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Towards Efficient Coarse-grained Dialogue Response Selection2
This work proposes an efficient coarse-grained response selection subsystem consisting of two novel methods, the Hierarchical Dense Retrieval and the Deep Semantic Hashing, which achieves over 5x FLOPs speedup and over 192x storage compression ratio.
- 2
- 2
- 2
- 2
- 2
- SciNER: A Novel Scientific Named Entity Recognizing Framework2
A novel scientific named entity recognizing pipeline framework, called SciNER, which outperforms state-of-the-art baselines on several datasets in both computer science and biomedical domains.
- 2
- Fast and Accurate Bilingual Lexicon Induction via Matching Optimization2
Empirical experiments demonstrate that the proposed BLIMO method outperforms state-of-the-arts baselines greatly on two standard benchmarks and converts the lexicon induction problem into a maximum weighted matching problem, which could be efficiently solved by the matching optimization with greedy search.
- 2
- 2
- Your Reasoning Model Knows What Counts: Self-Guided Chain-of-Thought Pruning for Efficient Reasoning1
- 1
- EXCEEDS: Extracting Complex Events via Nugget-based Grid Modeling in Scientific Domain1
This work proposes 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.
- MMWOZ: Building Multimodal Agent for Task-oriented Dialogue1
A novel multimodal model called MATE (Multimodal Agent for Task-oriEnted dialogue) is proposed as the baseline model for the MMWOZ dataset and comprehensive experimental analysis is conducted using MATE to investigate the construction of a practical multimodal agent for task-oriented dialogue.
- ESTJ: Enhancing Structured Tendency Judgment in Hybrid-Modal Table Understanding1
A modality preference alignment model, called ESTJ, is proposed, called ESTJ, which Enhances Structured Tendency Judgment in HMTU, which alleviates modality preference imbalance by adhering to the principle of least modality priority.
- 1
- 1
- 1
- Position Debiasing Fine-Tuning for Causal Perception in Long-Term Dialogue1
A novel method is proposed, named Causal Perception long-term Dialogue framework (CPD), which employs perturbation-based causal variable discovery method to extract casually relevant utterances from the dialogue history and enhances model causal perception during fine-tuning.
- 1
- 1
- CPLR-SFS: Contrastive Prompt Learning to Reduce Redundancy for Scientific Faceted Summarization1
This work proposes a novel Contrastive Prompt Learning method to Reduce redundancy for Scientific Faceted Summarization, named CPLR-SFS, to generate concise and less-overlapping faceted summaries.
- SNER-CS: Self-training Named Entity Recognition in Computer Science1
A novel self-training method based pretraining language model with a distantly supervised label automatic construction system in CS (SNER-CS) is proposed, which performs previous state-of-the-art methods in computer science NER task.
- Gated Mechanism Enhanced Multi-Task Learning for Dialog Routing1
This paper proposes a comprehensive and general solution with multi-task learning framework, specifically including a novel dialog encoder and two tailored gated mechanism modules that can play the role of hierarchical information filtering and is non-invasive to the existing dialog systems.
- SciMRC: Multi-perspective Scientific Machine Reading Comprehension1
A novel multi-perspective scientific machine reading comprehension dataset, SciMRC, which incorporates perspectives from beginners, students, and experts is introduced, which highlights the challenging nature of the scientific machine comprehension tasks.
- 1
- 1
- 1
- A Span-Based Distantly Supervised NER with Self-learning1
A novel span-based self-learning method, which employs span-level features to update corresponding dictionaries based on both local and global features, and performs better than the state-of-the-art baselines.
- 1
- 1
- Streamlined Decoder for Chinese Spoken Language Understanding1
A novel streamlined decoding structure with attention mechanism, which uses three successively connected RNN to decode act, slot and value respectively, which exceeds state-of-the-art model on an unaligned multi-turn task-oriented Chinese spoken dialogue dataset provided by the contest.
- 1
- –
- –
- –
- –
- G-CoS: An Interpretable Gain-Cost Framework for User Satisfaction Estimation in Generative Information Retrieval–
The Gain-aware Cost-sensitive Satisfaction estimator (G-CoS), an interpretable gain-cost framework for user satisfaction estimation in GenIR, which models user satisfaction as a dynamic trade-off between Response Quality and multidimensional Interaction Cost.
- –
- –
- A Distributed Collaborative Retrieval Framework Excelling in All Queries and Corpora based on Zero-shot Rank-Oriented Automatic Evaluation–
A novel Distributed Collaborative Retrieval Framework is proposed, outperforming each single model across all queries and corpora, and can achieve performance comparable to effective listwise methods like RankGPT and ListT5, while offering superior efficiency.
- –
- Measuring Cross-Lingual Transferability of Multilingual Transformers on Sentence Classification–
IGap, a cross-lingual transferability metric for multilingual Transformers on sentence classification tasks, is proposed, which takes training error into consideration, and can also estimate transferability without end-task data.
- CRAT-XML: Contrastive Representation Adversarial Training for Extremely Multi-Label Text Classification–
CRAT-XML is proposed, which uses contrast adversarial learning to optimize text representation and enhance the acquisition of dependency relations between text and labels, thus reducing the need for integration at the representation level and achieving relatively high accuracy under low-resource and low-time conditions.
- –
- –
- Incorporating Domain Knowledge into Text Classification Diagnosis in Customer Service Dialogue Field–
A domain knowledge-based text classification diagnosis method, which innovatively transforms the question and answer task into the text classification task and uses an ERNIE-based structure to match consumer questions with multivariate groups of answers from the knowledge base, judged by similarity.
- Customer Satisfaction Research based on Customer Service Dialogue Corpus–
This work introduces an ERNIE-based customer satisfaction analysis model that automatically analyses the text of customer service dialogues and scores them from four perspectives (i.e., product, service, process and overall) without human involvement.
- –
- –
- –
- –
- –
- Deep Kernel Supervised Hashing for Network Embedding–
A novel deep hashing method for signed social network embedding by considering simultaneously positive and negative links is proposed, which performs better than several state-of-the-art baselines through link prediction task over two real-world signed social networks.
- –
- Picture News Collection: A Dataset for Automatic Picture News Thumbnail Selection–
The automatic picture news thumbnail selection task is introduced, an attention-based thumbnail selection model is proposed, and experimental results comparing with three image classification based baselines show that the proposed methods outperform the baselines.
- –
- Enhancing Neural Sequence Labeling with Position-Aware Self-Attention.–
An innovative and well-designed attention-based model is proposed (called position-aware self-attention, i.e., PSA) within a neural network architecture, to explore the positional information of an input sequence for capturing the latent relations among tokens.
- –
- –
- A Comparative Study of Answer-Contained Snippets and Traditional Snippets–
This paper proposes a simple answer-contained snippet method for community-based Question and Answer (cQA) search, and compares the method with the state-of-the-art traditional snippet algorithms.
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.