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
- ChengXiang ZhaiSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works260 from public data
- TwitterRank2,100
Experimental results show that TwitterRank outperforms the one Twitter currently uses and other related algorithms, including the original PageRank and Topic-sensitive PageRank, which is proposed to measure the influence of users in Twitter.
- Comparing Twitter and Traditional Media Using Topic Models1,338
This paper empirically compare the content of Twitter with a traditional news medium, New York Times, using unsupervised topic modeling, and finds interesting and useful findings for downstream IR or DM applications.
- Instance Weighting for Domain Adaptation in NLP915
This paper formally analyze and characterize the domain adaptation problem from a distributional view, and shows that there are two distinct needs for adaptation, corresponding to the different distributions of instances and classification functions in the source and the target domains.
- 597
- Jointly modeling aspects, ratings and sentiments for movie recommendation (JMARS)464
A probabilistic model based on collaborative filtering and topic modeling is proposed that allows it to capture the interest distribution of users and the content distribution for movies; it provides a link between interest and relevance on a per-aspect basis and it allows us to differentiate between positive and negative sentiments on aPer-Aspect basis.
- Learning Natural Language Inference with LSTM461
A special long short-term memory (LSTM) architecture for NLI that remembers important mismatches that are critical for predicting the contradiction or the neutral relationship label and achieves an accuracy of 86.1%, outperforming the state of the art.
- 455
- Jointly Modeling Aspects and Opinions with a MaxEnt-LDA Hybrid452
This paper proposes a MaxEnt-LDA hybrid model to jointly discover both aspects and aspect-specific opinion words and shows that with a relatively small amount of training data, this model can effectively identify aspect and opinion words simultaneously.
- Adaptive filters for continuous queries over distributed data streams449
This work considers an environment where distributed data sources continuously stream updates to a centralized processor that monitors continuous queries over the distributed data, and proposes a new technique for reducing the overhead.
- 376
- Knowledge transfer via multiple model local structure mapping349
A locally weighted ensemble framework to combine multiple models for transfer learning, where the weights are dynamically assigned according to a model's predictive power on each test example, is proposed.
- 346
- Improving Multimodal Named Entity Recognition via Entity Span Detection with Unified Multimodal Transformer343
A Unified Multimodal Transformer to guide the final predictions with the entity span predictions and achieves the new state-of-the-art performance on two benchmark datasets.
- Adapting BERT for Target-Oriented Multimodal Sentiment Classification339
This work studies Target-oriented Multimodal Sentiment Classification (TMSC) and proposes a multimodal BERT architecture and borrows the idea from selfattention and design a target attention mechanism to perform target-image matching to derive targetsensitive visual representations.
- Federated Learning from Pre-Trained Models: A Contrastive Learning Approach322
This work designs a Federated Prototype-wise Contrastive Learning (FedPCL) approach which shares knowledge across clients through their class prototypes and builds client-specific representations in a prototype-wise contrastive manner, and performs a thorough evaluation of the proposed FedPCL in the lightweight framework.
- Improving Multi-hop Knowledge Base Question Answering by Learning Intermediate Supervision Signals319
A novel teacher-student approach for the multi-hop KBQA task, where the teacher network tries to learn intermediate supervision signals for improving the reasoning capacity of the student network, and both forward and backward reasoning are utilized to enhance the learning of intermediate entity distributions.
- R3: Reinforced Ranker-Reader for Open-Domain Question Answering305
This paper proposes a new pipeline for open-domain QA with a Ranker component, which learns to rank retrieved passages in terms of likelihood of extracting the ground-truth answer to a given question, and proposes a novel method that jointly trains the Ranker along with an answer-extraction Reader model, based on reinforcement learning
- Query Graph Generation for Answering Multi-hop Complex Questions from Knowledge Bases267
This paper proposes a modified staged query graph generation method with more flexible ways to generate query graphs that achieves the state of the art on three benchmark KBQA datasets.
- Topical Keyphrase Extraction from Twitter256
A context-sensitive topical PageRank method for keyword ranking and a probabilistic scoring function that considers both relevance and interestingness of keyphrases for keyphrase ranking are proposed.
- Finding Bursty Topics from Microblogs253
A topic model that simultaneously captures two observations is proposed that helps find event-driven posts on microblogs and helps identify and filter out "personal" posts.
- Interpretable Rumor Detection in Microblogs by Attending to User Interactions245
This work proposes a post-level attention model (PLAN) to model long distance interactions between tweets with the multi-head attention mechanism in a transformer network and is the first to evaluate the models on two rumor detection data sets: the PHEME data set as well as the Twitter15 and Twitter16 data sets.
- 233
- Learning Sentence Embeddings with Auxiliary Tasks for Cross-Domain Sentiment Classification224
This paper borrows the idea from Structural Correspondence Learning and uses two auxiliary tasks to help induce a sentence embedding that supposedly works well across domains for sentiment classification and proposes to jointly learn this sentenceembedding together with the sentiment classI-class itself.
- CQArank224
This work proposed Topic Expertise Model (TEM), a novel probabilistic generative model with GMM hybrid, to jointly model topics and expertise by integrating textual content model and link structure analysis, and proposed CQARank to measure user interests and expertise score under different topics.
- A Survey on Complex Knowledge Base Question Answering: Methods, Challenges and Solutions223
This paper elaborately summarize the typical challenges and solutions for complex KBQA, and presents the two mainstream categories of methods, namely semantic parsing-based (SP-based) methods and information retrieval-based [IR-based] methods.
- A Systematic Exploration of the Feature Space for Relation Extraction211
This paper systematically explore a large space of features for relation extraction and evaluates the effectiveness of different feature subspaces, and presents a general definition of feature spaces based on a graphic representation of relation instances.
- TopicSketch: Real-Time Bursty Topic Detection from Twitter210
This paper proposes a sketch-based topic model together with a set of techniques to achieve real-time detection of bursty topics in Twitter, and demonstrates both efficiency and effectiveness of the approach.
- 199
- RegMix: Data Mixture as Regression for Language Model Pre-training196
RegMix is proposed to automatically identify a high-performing data mixture by formulating it as a regression task and consistently outperforms human selection in experiments involving models up to 7B models trained on 100B tokens, while matching or exceeding DoReMi using just 10% of the computational resources.
- Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast182
The results show that feeding an (infectious) adversarial image into the memory of any randomly chosen agent is sufficient to achieve infectious jailbreak, and derive a simple principle for determining whether a defense mechanism can provably restrain the spread of infectious jailbreak.
- Information Extraction from Text180
This chapter provides a survey of the major work on named entity recognition and relation extraction in the past few decades, with a focus on work from the natural language processing community.
- 166
- Complex Knowledge Base Question Answering: A Survey162
A review of recent advances in KBQA with the focus on solving complex questions, which usually contain multiple subjects, express compound relations, or involve numerical operations, and two mainstream categories of methods, namely semantic parsing- based (SP-based) methods and information retrieval-based (IR) methods.
- 153
- Disentangling Hate in Online Memes127
The proposed DisMultiHate is designed to disentangle target entities in multimodal memes to improve the hateful content classification and explainability and is able to outperform state-of-the-art unimodal and multi-modal baselines in the hateful meme classification task.
- Recurrent Neural Networks with Auxiliary Labels for Cross-Domain Opinion Target Extraction117
This work uses rule-based unsupervised methods to create auxiliary labels and use neural network models to learn a hidden representation that works well for different domains and finds that it can outperform a number of strong baselines with a large margin.
- A two-stage approach to domain adaptation for statistical classifiers117
This paper presents a two-stage approach to domain adaptation, where at the first generalization stage, the author looks for a set of features generalizable across domains, and at the second adaptation stage, heuristics to pick up useful features specific to the target domain.
- Pro-Cap: Leveraging a Frozen Vision-Language Model for Hateful Meme Detection115
This work proposes a probing-based captioning approach to leverage pre-trained vision-language models in a zero-shot visual question answering (VQA) manner and prompts a frozen PVLM by asking hateful content-related questions and uses the answers as image captions (which it calls Pro-Cap), so that the captions contain information critical for hateful content detection.
- A Co-Matching Model for Multi-choice Reading Comprehension109
A new co-matching approach to this problem, which jointly models whether a passage can match both a question and a candidate answer, achieves state-of-the-art performance.
- Global Inference for Aspect and Opinion Terms Co-Extraction Based on Multi-Task Neural Networks108
This work first applies a multi-task learning framework to implicitly capture the relations between the two tasks, and proposes a global inference method by explicitly modelling several syntactic constraints among aspect term extraction and opinion term extraction to uncover their intra-task and inter-task relationship.
- Modelling Domain Relationships for Transfer Learning on Retrieval-based Question Answering Systems in E-commerce106
This paper proposes a general framework, which can effectively and efficiently adapt the shared knowledge learned from a resource-rich source domain to aresource-poor target domain and proposes an efficient and effective hybrid model by combining a sentence encoding- based method and a sentence interaction-based method as the base model.
- Linking Entities to a Knowledge Base with Query Expansion106
A novel approach to entity linking based on a statistical language model-based information retrieval with query expansion with a strong emphasis on named entities in the local contexts and a positional language model to weigh them differently based on their distances to the query.
- Improving Multi-label Emotion Classification via Sentiment Classification with Dual Attention Transfer Network92
A new transfer learning architecture is proposed to divide the sentence representation into two different feature spaces, which are expected to respectively capture the general sentiment words and the other important emotion-specific words via a dual attention mechanism.
- 87
- 84
- Multi-task transfer learning for weakly-supervised relation extraction84
This paper proposes to use a multi-task learning method coupled with human guidance to address the weakly-supervised relation extraction problem, and proposes a framework that models the commonality among different relation types through a shared weight vector.
- 79
- Prompt Federated Learning for Weather Forecasting: Toward Foundation Models on Meteorological Data78
A novel federated learning approach has been proposed to collaboratively learn a brand-new spatio-temporal Transformer-based foundation model across participants with heterogeneous meteorological data to relieve the data exposure concern across regions.
- Hashtag recommendation with topical attention-based LSTM78
Experimental results show that the model significantly outperforms various competitive baseline methods and the incorporation of topical attention mechanism gives more than 7.4% improvement in F1 score compared with standard LSTM method.
- An empirical study of tokenization strategies for biomedical information retrieval77
It is shown that different query types require different tokenization heuristics, stemming is effective only for certain queries, and stop word removal in general does not improve the retrieval performance on biomedical text.
- Using Syntactic and Semantic Relation Analysis in Question Answering.75
In TREC-13, the use of analysis to extract answer n and list questions from exter These external answer nugg directly projected back to th questions) or be used as add evidence for corpus-based (for factoid questions) is proposed.
- Generating Templates of Entity Summaries with an Entity-Aspect Model and Pattern Mining61
A novel approach to automatic generation of summary templates from given collections of summary articles by developing an entity-aspect LDA model that automatically grouping of semantically related sentence patterns and automatic identification of template slots that need to be filled in.
- 60
- 58
- It Is Not Just What We Say, But How We Say Them: LDA-based Behavior-Topic Model58
Textual information exchanged among users on online social network platforms provides deep understanding into users’ interest and behavioral patterns, which has not yet been well incorporated in the existing topic modeling.
- Intriguing Properties of Data Attribution on Diffusion Models57
This work presents a significantly more efficient approach for attributing diffusion models, while the unexpected findings suggest that at least in non-convex settings, constructions guided by theoretical assumptions may lead to inferior attribution performance.
- Knowledge Base Question Answering with Topic Units57
This paper proposes to perform topic unit linking where topic units cover a wider range of units of a KB, and uses a generation-and-scoring approach to gradually refine the set of topic units.
- Knowledge Base Question Answering With a Matching-Aggregation Model and Question-Specific Contextual Relations56
This paper explores the use of a “matching-aggregation” framework to match candidate answers with questions and makes use of question-specific contextual relations to enhance the representations of candidate answer entities.
- Generating Aspect-oriented Multi-Document Summarization with Event-aspect model56
Key features of this method include automatic grouping of semantically related sentences and sentence ranking based on extension of random walk model and a new sentence compression algorithm which use dependency tree instead of parser tree.
- Exploiting domain structure for named entity recognition55
A simple yet effective way to automatically rank features based on their generalizabilities across domains and a domain-aware cross validation strategy to help choose an appropriate parameter for the rank-based prior is proposed.
- 53
- Real Time Event Detection in Twitter52
A mixture Gaussian model for bursty word extraction in Twitter and then a novel time-dependent HDP model for new topic detection, which can grasp new events, the location and the time an event becomes bursty promptly and accurately are proposed.
- 51
- Detecting Personal Intake of Medicine from Twitter51
A classifier for identifying mentions of personal intake of medicine in tweets is developed using a stacked ensemble of shallow convolutional neural network models on an annotated dataset and it is believed that the developed classifier has direct uses in the areas of psychology, health informatics, pharmacovigilance, and affective computing.
- Coupled Hierarchical Transformer for Stance-Aware Rumor Verification in Social Media Conversations50
To extend BERT to obtain thread representations, this paper proposes a Hierarchical Transformer, which divides each long thread into shorter subthreads, and employs BERT to separately represent each subthread, followed by a global Transformer layer to encode all the subthreads.
- Causal Reinforcement Learning: A Survey49
This survey comprehensively review the literature on causal reinforcement learning and introduces the basic concepts of causality and reinforcement learning, and explains how causality can address core challenges in non-causal reinforcement learning.
- Domain adaptation in natural language processing48
The instance weighting and feature selection solutions are formally developed into two general and principled frameworks for domain adaptation, which modify the objective function of the standard risk minimization framework for supervised learning, and include standard supervised learning and semi-supervised learning as special cases.
- Generating gene summaries from biomedical literature: A study of semi-structured summarization47
A study of methods for automatically generating gene summaries from biomedical literature which consists of sentences covering specific semantic aspects of a gene and a two-stage approach to generate such a summary for a given gene.
- 46
- Who is Retweeting the Tweeters? Modeling, Originating, and Promoting Behaviors in the Twitter Network46
This article uses retweeting activities among Twitter users to define and model originating and promoting behavior, and proposes a basic model for measuring the two behaviors, a mutual dependency model, and a range-based model, which considers the depth and reach of users’ original tweets.
- Can Syntax Help? Improving an LSTM-based Sentence Compression Model for New Domains45
This paper proposes two major changes to the LSTM neural network model for sentence compression: using explicit syntactic features and introducing syntactic constraints through Integer Linear Programming (ILP).
- 44
- Topical Co-Attention Networks for hashtag recommendation on microblogs44
A novel Topical Co-Attention Network (TCAN) is proposed that jointly models content attention and topic attention simultaneously, in the sense that the content representation(s) are used to guide the topic attention and the topic representation is used to Guide content attention.
- Learning Topics and Positions from Debatepedia44
A generative model positing latent topics and cross-cutting positions that gives special treatment to person mentions and opinion words is introduced that’s usefulness in attaching opinionated documents to arguments and its consistency with human judgments about positions is evaluated.
- Multi-hop Knowledge Base Question Answering with an Iterative Sequence Matching Model41
This paper proposes a novel iterative sequence matching model to address several limitations of previous methods for multi-hop KBQA and demonstrates that it can significantly outperform existing methods on three different benchmark datasets.
- Extracting paraphrases of technical terms from noisy parallel software corpora39
This paper proposes a number of techniques to address the noisy data problem of extracting technical paraphrases from a parallel software corpus, namely, a collection of duplicate bug reports.
- Modularized Networks for Few-shot Hateful Meme Detection38
This paper addresses the challenge of detecting hateful memes in the low-resource setting where only a few labeled examples are available by fine-tuning large language models with LoRA on selected tasks pertinent to hateful meme detection, thereby generating a suite of LoRA modules.
- Investigating Math Word Problems using Pretrained Multilingual Language Models36
The experiments show that the MWP solvers may not be transferred to a different language even if the target expressions share the same numerical constants and operator set, and it can be better generalized if problem types exist on both source language and target language.
- 36
- Modeling User Arguments, Interactions, and Attributes for Stance Prediction in Online Debate Forums35
This work proposes an integrated model to leverage side information, a regression-based latent factor model which jointly models user arguments, interactions, and attributes and demonstrates in experiments that the method has promising results on both micro-level and macro-level stance prediction.
- A Latent Variable Model for Viewpoint Discovery from Threaded Forum Posts34
A principled generative latent variable model which captures three important factors: viewpoint specific topic preference, user identity and user interactions is proposed which clearly outperforms a number of baseline models in terms of both clustering posts based on viewpoints and clustering users with different viewpoints.
- Escape from An Echo Chamber33
A novel news-analysis platform that provides a panoramic view of stances towards a particular event from different news media sources and adopts a news stance classification model to help users escape from an echo chamber.
- Predicting User’s Political Party Using Ideological Stances32
This work exploits users' ideological stances on controversial issues to predict political party of online users and proposes a collaborative filtering approach to solve the data sparsity problem of users stances on ideological topics and applies clustering method to group the users with the same party.
- Extracting Relation Descriptors with Conditional Random Fields32
This paper proposes two modifications to linear-chain CRFs, namely, reducing the space of possible label sequences and introducing long-range features, and finds that both modifications can significantly improve the performance for this novel relation extraction problem.
- 31
- AUTOMATICALLY GENERATING GENE SUMMARIES FROM BIOMEDICAL LITERATURE31
The software, the first that automatically generates gene summaries from biomedical literature, is described, which involves first retrieving relevant articles and then extracting the most informative sentences from the retrieved articles to generate a structured gene summary.
- Do You Trust to Get Trust? A Study of Trust Reciprocity Behaviors and Reciprocal Trust Prediction30
This paper seeks to determine the extent to which reciprocity exists in a trust network and develops quantitative models for measuring reciprocity and reciprocity related behaviors, and identifies several reciprocity behaviors and their respective measures.
- Extraction of coherent relevant passages using hidden Markov models30
This article presents a new method for accurately detecting coherent relevant passages of variable lengths using hidden Markov models (HMMs) and shows that with appropriate parameter estimation, the HMM method outperforms a number of strong baseline methods on two datasets.
- Synergizing Large Language Models and Pre-Trained Smaller Models for Conversational Intent Discovery29
This work proposes SynCID, which harnesses the profound semantic comprehension of LLMs alongside the operational agility of SLMs for CID, and leverages the in-context learning strengths of LLMs to generate labels for new intents.
- Modeling Transitions of Focal Entities for Conversational Knowledge Base Question Answering29
A novel graph-based model is proposed to capture the transitions of focal entities and apply a graph neural network to derive a probability distribution of focal individuals for each question, which is then combined with a standard KBQA module to perform answer ranking.
- Multi-source domain adaptation for panoramic semantic segmentation28
A novel framework, Deformation Transform Aligner for Panoramic Semantic Segmentation (DTA4PASS), which converts all pinhole images in the source domains into distorted images and aligns the source distorted and panoramic images with the target panoramic images.
- Does Continual Learning Equally Forget All Parameters?27
This paper studies which modules in neural networks are more prone to forgetting by investigating their training dynamics during continual learning (CL), and finds that finetuning them only on a small buffer at the end of any CL method can bring non-trivial improvement.
- 25
- Federated Prompt Learning for Weather Foundation Models on Devices25
This paper proposes Federated Prompt learning for Weather Foundation Models on Devices (FedPoD), which enables devices to obtain highly customized models while maintaining communication efficiency and leads the performance among state-of-the-art baselines across various setting in real-world on-device weather forecasting datasets.
- FedNoiL: A Simple Two-Level Sampling Method for Federated Learning with Noisy Labels25
A simple two-level sampling method that selects clean labels and correct pseudo-labels at the client end for more robust local training and investigates different schedules changing the local epochs between aggregations over the course of FL, which notably improves the communication and computation efficiency in noisy label setting.
- Does BERT Understand Idioms? A Probing-Based Empirical Study of BERT Encodings of Idioms25
The experiment results suggest that BERT indeed can separate the literal and idiomatic usages of a PIE with high accuracy and it is also able to encode the idiomatic meaning of an idiomatic expression to some extent.
- Personalized Microtopic Recommendation on Microblogs25
This article proposes a joint probabilistic latent factor model to integrate rich information into a matrix factorization-based solution to microtopic recommendation and shows that this model significantly outperforms a few competitive baseline methods, especially in the circumstance where users have few adoption behaviors.
- A Unified Model for Topics, Events and Users on Twitter25
A model which combines an LDA-like topic model and the Recurrent Chinese Restaurant Process to capture topics and events on Twitter is proposed and can accurately identify meaningful events and the event-topic affinity vectors are effective for event recommendation and grouping events by topics.
- Modeling interaction features for debate side clustering25
A two-stage solution based on latent variable models is proposed: an interaction feature identification stage to mine interaction features from structured debate posts with known sides and reply intentions; and a clustering stage to incorporate interaction features and model the interplay between interactions and sides for debate side clustering.
- Identifying Event-related Bursts via Social Media Activities25
This paper proposes to identify event-related bursts via social media activities by studying how to correlate multiple types of activities to derive a global bursty pattern, and proposes a novel function which can capture the state context.
- Federated Low-Rank Adaptation for Foundation Models: A Survey24
This survey examines how LoRA has been integrated into federated fine-tuning for foundation models—an area known as FedLoRA by focusing on three key challenges: distributed learning, heterogeneity, and efficiency.
- 24
- 24
- Topic modeling with document relative similarities23
A general model that links LDA with constraints derived from document relative similarities, and the constraints act as a regularizer of the log likelihood of LDA.
- Your love is public now: questioning the use of personal information in authentication22
It is demonstrated that simply liking a Facebook page does not corroborate that the user is interested in the page, and how this inferred information can be used to break a preference based backup authentication system is shown.
- NOAHQA: Numerical Reasoning with Interpretable Graph Question Answering Dataset20
NoAHQA is introduced, a conversational and bilingual QA dataset with questions requiring numerical reasoning with compound mathematical expressions and a new QA model for generating a reasoning graph where the reasoning graph metric still has a large gap compared with that of humans.
- Cross-Thought for Sentence Encoder Pre-training20
This paper proposes Cross-Thought, a novel approach to pre-training sequence encoder, which is instrumental in building reusable sequence embeddings for large-scale NLP tasks such as question answering and textual entailment tasks.
- Finding Thoughtful Comments from Social Media20
This paper empirically study various textual features, discourse relations and relevance features, along with basic textual features to predict thoughtful comments to predict the comment quality in terms of thoughtfulness and found that the most useful features include the discourse relationsand relevance features.
- On Modeling Virality of Twitter Content20
This paper identifies the most viral messages in GE2011 as well as the users behind them, and introduces several quantitative indices for measuring the virality of tweets that are retweeted.
- Recurrent Chinese Restaurant Process with a Duration-based Discount for Event Identification from Twitter19
A new nonparametric model which considers burstiness is proposed which is based on the Recurrent Chinese Restaurant Process and combined with traditional topic models to identify both events and topics simultaneously on Twitter.
- Joint Learning for Coreference Resolution with Markov Logic19
This work proposes a joint learning model which combines pairwise classification and mention clustering with Markov logic and shows competitive performance against the best system from CoNLL-2011.
- UIUC in HARD 2004--Passage Retrieval Using HMMs.19
The results show that the HMM method can improve the boundaries of pre-segmented passages in terms of overall passage retrieval accuracy and recall, but at the price of precision sometimes.
- Enhancing Document-level Translation of Large Language Model via Translation Mixed-instructions18
The proposed translation mixed-instructions enable LLMs (Llama-2~7B and 13B) to maintain consistent translation performance from the sentence level to documents containing as many as 2048 tokens and significantly improves translation quality, both in terms of BLEU score and discourse coherence.
- Continual Task Allocation in Meta-Policy Network via Sparse Prompting18
Continual Task Allocation via Sparse Prompting (CoTASP), which learns over-complete dictionaries to produce sparse masks as prompts extracting a sub-network for each task from a meta-policy network, achieves a promising plasticity-stability trade-off without storing or replaying any past tasks' experiences.
- 18
- Aspect and opinion aware abstractive review summarization with reinforced hard typed decoder17
A two-stage reinforcement learning approach is proposed, which first predicts the output word type from the three types, and then leverages the predicted word type to generate the final word distribution.
- Extracting Interest Tags from Twitter User Biographies17
This work study how to extract information revealing users’ personal interests from Twitter biographies, and finds that there is a weak linkage between them, suggesting that biographies can potentially serve as a complimentary source of information to tweets.
- Unsupervised Deep Structured Semantic Models for Commonsense Reasoning15
Two neural network models based on the Deep Structured Semantic Models (DSSM) framework are proposed to tackle two classic commonsense reasoning tasks, Winograd Schema challenges (WSC) and Pronoun Disambiguation (PDP).
- Lifetime Lexical Variation in Social Media15
A latent variable model that jointly models the lexical content of tweets and Twitter users’ ages and can learn meaningful age-specific topics such as “school” for teenagers and “health’ for older people is presented.
- Cross-Topic Rumor Detection using Topic-Mixtures14
This paper proposes a method inspired by the “mixture of experts” paradigm that can outperform two baseline debiasing methods in a cross-topic setting and works better than the baselines, showing that the method does not rely on superficial features.
- Towards Opinion Summarization from Online Forums14
This work constructs an annotated corpus to facilitate the study of extractive summarization of online discussion forums, and defines a set of features to capture relevance, text quality and subjectivity, and empirically test their usefulness in choosing summary sentences.
- Mining User Relations from Online Discussions using Sentiment Analysis and Probabilistic Matrix Factorization14
Experiments with two tasks show that the learned latent factor representation can give good performance on a relation polarity prediction task and improve the performance of a subgroup detection task.
- A Neural Network Model for Semi-supervised Review Aspect Identification13
A neural network model to identify aspects from reviews by learning their distributional vectors is proposed, which can outperform a few baselines in terms of aspect quality, perplexity and sentence clustering results.
- 13
- 12
- ngram-OAXE: Phrase-Based Order-Agnostic Cross Entropy for Non-Autoregressive Machine Translation12
This work extends oaxe by only allowing reordering between ngram phrases and still requiring a strict match of word order within the phrases, and shows that ngram noaxe indeed improves the translation of ngram words, and produces more fluent translation with a better modeling of sentence structure.
- 12
- 12
- 11
- Translate-Train Embracing Translationese Artifacts11
It is found that artifacts have common patterns in different languages and can be modeled by deep learning, and subsequently proposed an approach to conduct translate-train using Translationese Embracing the effect of Artifacts (TEA), which outperforms strong baselines.
- VICTOR: An Implicit Approach to Mitigate Misinformation via Continuous Verification Reading10
VICTOR, an easy-to-apply module on top of a recommender system to mitigate misinformation, takes an elegant, implicit approach to deliver fake-news verifications, such that readers of fake news can continuously access more verified news articles aboutfake-news events without explicit correction.
- 10
- A BERT-Based Two-Stage Model for Chinese Chengyu Recommendation10
A two-stage model is proposed, where during the first stage the authors re-train a Chinese BERT model by masking out Chengyu from a large Chinese corpus with a wide coverage of Chengyu, and during the second stage, the re-trained, Chengyu-oriented BERT on a specific Chengyu recommendation dataset is fine-tuned.
- All the Wiser10
A news reading platform is developed in which reinforcement learning is leveraged to learn an intervention module on top of a recommender system such that the module is activated to replace RS to recommend news toward the verification once users touch the fake news.
- 10
- Learning and Evaluating Chinese Idiom Embeddings10
A BERT-based method is presented that directly learns embedding vectors for individual idioms in Chinese idiom embeddings and substantially outperforms existing methods on the evaluation dataset.
- A Hassle-Free Unsupervised Domain Adaptation Method Using Instance Similarity Features10
This work presents a simple yet effective unsupervised domain adaptation method that can be generally applied for different NLP tasks, and shows that this method consistently outperforms a few baselines, including SCL, an existing general un superviseddomain adaptation method widely used in NLP.
- Improving Compositional Generalization in Math Word Problem Solving9
This manuscript introduces a data splitting method to create compositional splits from existing MWP datasets and proposes an iterative data augmentation method that includes diverse compositional variation into training data and could collaborate with MWP methods.
- Unsupervised Video Domain Adaptation for Action Recognition: A Disentanglement Perspective9
This work tackles unsupervised video domain adaptation from a disentanglement view, and proposes several objectives to constrain the latent factors in TranSVAE to better serve for adaptation.
- COSY: COunterfactual SYntax for Cross-Lingual Understanding9
This work includes the design of SYntax-aware networks as well as a COunterfactual training method to implicitly force the networks to learn not only the semantics but also the syntax, based on the observation that universal syntax is transferable across different languages.
- Query-Oriented Keyphrase Extraction9
The empirical evaluation shows that using a named entity-based phraseness criterion and a language model-based informativeness score gives the best performance on the authors' task, and this method also outperforms two state-of-the-art baseline methods.
- Engagingness and Responsiveness Behavior Models on the Enron Email Network and Its Application to Email Reply Order Prediction9
This paper investigates user engagingness and responsiveness as two interaction behaviors that give us useful insights into how users email one another and proposes four types of models to quantify engagings and responsiveness of users.
- Unsupervised Information Extraction with Distributional Prior Knowledge9
A generative model that incorporates distributional prior knowledge to help distribute candidates in a document into appropriate slots is proposed that can bring substantial improvements to the task of automatic discovery of information extraction template from a given text collection.
- 9
- Do Existing Testing Tools Really Uncover Gender Bias in Text-to-Image Models?8
An enhanced detector called CLIP-Enhance is proposed, which most accurately measures the gender bias in T2I models, with a difference of only 0.47%-1.23%, and most effectively filters out 82.91% of low-quality images.
- A fast interpolation-based multi-objective evolutionary algorithm for large-scale multi-objective optimization problems8
A fast interpolation-based multi-objective evolutionary algorithm for solving large-scale multi-objective optimization problems with high convergence speed and accuracy and the superiority of the proposed algorithm over other state-of-the-art algorithms is verified.
- 8
- Multi-Level Head-Wise Match and Aggregation in Transformer for Textual Sequence Matching8
Experiments show that the proposed new approach to sequence pair matching with Transformer can achieve new state-of-the-art performance on multiple tasks that rely only on pre-computed sequence-vector-representation, such as SNLI, MNLI-match,MNLI-mismatch, QQP, and SQuAD-binary.
- A BERT-based Dual Embedding Model for Chinese Idiom Prediction8
A BERT-based dual embedding model is proposed to encode the contextual words as well as to learn dual embeddings of the idioms to improve the performance of the Chinese idiom prediction task.
- Context modeling for ranking and tagging bursty features in text streams8
Using a large corpus of a stream of news articles, it is quantitatively shown that the proposed context language models for bursty features can effectively help rank Bursty features based on their newsworthiness and to assign meaningful tags to annotatebursty features.
- SMU-SIS at TAC 2010 - KBP Track Entity Linking8
This paper proposes the two way entity linking approach to reformulate query, disambiguate the entity and link to the relevant KB repository, and developed several entity linking engines to evaluate the solution.
- Federated Vision-Language-Recommendation with Personalized Fusion7
FedMR uses a foundation model on the server to encode multimodal item data, and introduces a Mixing Feature Fusion Module on each client, which adjusts fusion strategy weights based on user interaction history to generate personalized item representations that capture users’ fine-grained preferences.
- Leveraging auxiliary tasks for document-level cross-domain sentiment classification7
This paper designs a new auxiliary task based on sentiment scores of domain-independent words, and proposes two neural network architectures to respectively induce document embeddings and sentence embeddeddings that work well for different domains.
- 6
- 6
- 6
- 6
- Personalized Microtopic Recommendation with Rich Information6
A novel hierarchical Bayesian model integrating user adoption behaviors, user item content information, and rich contextual information into the same principled model is proposed and Experimental results show that the model significantly outperforms a few baseline methods.
- An Integrated Model for User Attribute Discovery: A Case Study on Political Affiliation Identification6
This work proposes an integrated unsupervised model which takes into consideration all the three components integral to the task of discovering user demographic attributes from social media, and incorporates collaborative filtering with probabilistic matrix factorization to solve the data sparsity problem.
- A study of age gaps between online friends6
A strong hypothesis based on homophily is validated and adapted to ensure the certainty of user attribute extracted via weakly supervised propagation and a propagation-based algorithm is designed to predict online users' age, leveraging the clique-based hypothesis.
- 6
- Automatically building templates for entity summary construction6
Key features of this novel approach to automatic generation of summary templates from given collections of summary articles include automatic grouping of semantically related sentence patterns and automatic identification of template slots that need to be filled in.
- UDEL/SMU at TREC 2009 Entity Track6
The general goal is to study how to apply language modeling approaches and natural language processing techniques to the task and proposed to support information based on segment retrieval, to extract entities using Stanford NER tagger, and to rank entities based on a previously proposed probabilistic framework for expertding.
- 5
- GPT-Lab: Next Generation Of Optimal Chemistry Discovery By GPT Driven Robotic Lab5
This work introduces GPT-Lab, a paradigm that employs GPT models to give robots human-like intelligence in robotic experimentation, and mines literature for materials and methods and validates findings through high-throughput synthesis.
- 5
- An LSTM model for cloze-style machine comprehension5
An LSTM-based model designed earlier for textual entailment is adopted and two new models for cloze-style machine comprehension are proposed, both of which are quite competitive compared with the state of the art.
- Sentence Compression with Reinforcement Learning5
This work forms this task in a reinforcement learning framework, where hard constraints are used as rewards in a soft manner and achieves competitive performance with a large improvement on the speed.
- Inferring User Consumption Preferences from Social Media5
This paper proposes a principled latent variable model to infer user consumption preferences at the category level (e.g. inferring what categories of products a user would like to buy) and demonstrates how this model naturally links users’ published content and following relations on microblogs with their consumption behaviors on e-commerce websites.
- Pairwise relation classification with mirror instances and a combined convolutional neural network5
A neural network based method for relation classification that combines the raw sequence and the shortest dependency path representations of relation instances and uses mirror instances to perform pairwise relation classification is proposed.
- 5
- 4
- What Hides behind Unfairness? Exploring Dynamics Fairness in Reinforcement Learning4
This paper introduces a novel notion called dynamics fairness, which explicitly captures the inequality stemming from environmental dynamics, distinguishing it from those induced by decision-making or inherited from the past, and derives identification formulas that allow for reliable estimations from data.
- Interventional Training for Out-Of-Distribution Natural Language Understanding4
This paper proposes a novel interventional training method called Bottom-up Automatic Intervention (BAI) that performs multi-granular intervention with identified multifactorial confounders and shows the effectiveness of BAI for tackling OOD settings.
- MOHE-NET: MONOCULAR OBJECT HEIGHT ESTIMATION NETWORK USING DEEP LEARNING AND SCENE GEOMETRY4
The proposed Monocular Object Height Estimation Network (MOHE-Net) that includes a cascade of two networks that performs state-of-the-art and can be deployed for obstacle avoidance on autonomous platforms.
- 4
- Extracting and Normalizing Entity-Actions from Users' Comments4
This work defines a new problem in this line which is extracting entity-actionable knowledge from the users’ comments and proposes a principled approach to solve this problem and detect exactly matched entities with 75.1% F-score and exactly matched actions with 76.43%F-score.
- Link Type Based Pre-Cluster Pair Model for Coreference Resolution4
This paper's participation in the CoNLL-2011 shared task, Modeling Unrestricted Coreference in OntoNotes, is presented and a link type based pre-cluster pair model is proposed, which helps to determine whether two mentions in a piece of text refer to the same entity.
- 3
- 3
- Comparative evaluation and analysis of DNA N4-methylcytosine methylation sites using deep learning3
A variety of state-of-the-art deep learning models are compared on six benchmark datasets to evaluate their performance in 4mC methylation site detection and it is concluded that deep learning can greatly expand the potential ofmethylation site prediction.
- Personalized Federated Learning With Graph3
A novel structured federated learning framework to simultaneously learn the global model and personalized model using each client’s local relations with others and its private dataset and the effectiveness of the proposed method has been demonstrated in varying degrees of data non-iid settings.
- 3
- 3
- Modeling Social Media Content with Word Vectors for Recommendation3
This work applies vector representation of words to measure the semantic similarity between text to design a model that seamlessly integrates word vectors into a joint model of user feedback and text content.
- Generating Supplementary Travel Guides from Social Media3
This paper proposes to use a latent variable model to align forum threads with the section structure of well-written travel guides and modify an ILP-based summarization method to generate section-specific summaries.
- 3
- 3
- BeeSpace Navigator: exploratory analysis of gene function using semantic indexing of biological literature3
BeeSpace Navigator is a prototype software for exploratory analysis of gene function using biological literature that supports an automatic analogue of the curator process to extract functions, with a simple interface intended for all biologists.
- 3
- 3
- Accurately extracting coherent relevant passages using hidden Markov models3
It is shown that with appropriate parameter estimation, the HMM method outperforms a number of strong baseline methods on two data sets.
- 2
- 2
- Revisiting the Markov Property for Machine Translation2
It is indicated that MAT with an order larger than 4 can generate translations with quality on par with that of conventional autoregressive transformers.
- 2
- 2
- 2
- Context Modeling with Evidence Filter for Multiple Choice Question Answering2
This work proposes a simple yet effective approach termed evidence filtering to model the relationships between the encoded contexts with respect to different options collectively, and to potentially highlight the evidence sentences and filter out unrelated sentences.
- MBS-NET: A MOVING-CAMERA BACKGROUND SUBTRACTION NETWORK FOR AUTONOMOUS DRIVING2
This paper is the first to propose background subtraction for moving cameras using deep learning using specialized deep learning approach, the Moving-camera Background Subtraction Network (MBS-Net), which is robust to detect changing background in various scenarios and does not require training on foreground objects.
- Embedding WordNet Knowledge for Textual Entailment2
This paper embeds the lexical entailment knowledge contained in WordNet in specially-learned word vectors, which they are called “entailment vectors,” and incorporates these entailment vectors into a decomposable attention model for textual entailment.
- When a Friend Online is More Than a Friend in Life: Intimate Relationship Prediction in Microblogs2
This study focuses on using Sina Weibo to discover intimate offline relationships among users and introduces deep learning approaches to this relationship identity problem and adopts an integrated model to capture features from both user profile and mention message.
- myDeal: A Mobile Shopping Assistant Matching User Preferences to Promotions2
This paper presents myDeal, a system that automatically ranks deals according to user preferences and presents them to the user on their mobile device.
- A constrained multi-objective evolutionary algorithm based on constraint-diversity dominance sorting1
- FOCUS: Evaluating Pre-trained Vision-Language Models on Underspecification Reasoning1
A novel probing dataset called FO-CUS is introduced to evaluate whether state-of-the-art VLMs have the ability to handle underspecified ambiguous statements and reveals that VLMs still fall short in handling underspecification even when visual inputs that can help resolve the ambiguities are available.
- Colloquial Singaporean English Style Transfer with Fine-Grained Explainable Control1
This work constructs a large, high-quality dataset of formal and informal sentences, annotated across six linguistic aspects, and proposes a novel multi-agent framework where large language models (LLMs) act as expert agents for each linguistic aspect.
- 1
- Converting Vocal Performances into Sheet Music Leveraging Large Language Models1
This study addresses the challenge of automating the conversion of vocal performances into sheet music using a combination of novel techniques, including large language models (LLMs), and presents a method that successfully translates vocal audio input into display-ready sheet music.
- Voting from Nearest Tasks: Meta-Vote Pruning of Pre-trained Models for Downstream Tasks1
A simple but effective ''Meta-Vote Pruning (MVP)'' method that significantly reduces the pruning iterations for a new task by initializing a sub-network from the pruned models of its nearest tasks.
- 1
- RGCN-Based Multi-Label Facial Expression Recognition by Combining Semantic and Geometric Information1
- 1
- 1
- 1
- 1
- A Joint Model of Product Properties, Aspects and Ratings for Online Reviews1
This work proposes a new model on top of HFT to separate product properties and aspects, which explicitly separates the two types of latent factors but links both to product ratings.
- A Unified Topic-Style Model for Online Discussions1
A unified probabilistic graphical model is designed to capture both topic-driven words and styledriven words that can help to separate serious and unserious posts/users and identify slang words.
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- Bi-level Personalization for Federated Foundation Models: A Task-vector Aggregation Approach–
A bi-level personalization framework for federated fine-tuning on foundation models that can gain group-wise personalization information while mitigating the disturbance of irrelevant or interest-conflict clients with non-IID data is proposed.
- Improving SpO 2 estimation accuracy under varying respiratory intensities using whale optimization and Gaussian process regression–
Experimental results demonstrated that the WOA-optimized GPR (WOA-GPR) model exhibited superior performance across varying respiratory conditions, and validation using real-world physiological data reinforced its reliability for practical applications.
- –
- –
- –
- –
- –
- –
- Multimodal tag synset induction–
This paper presents an effective multi-modal solution for tag synsets induction task by leveraging the massive image tags pairs in image-centric social networks, such as Instagram and Pinterest.
- Coordinated and specific restricted Boltzmann machine for cross-modal retrieval–
A coordinated and specific restricted Boltzmann machine (a.k.a. CSRBM) that can distinguish the common part from modality-specific part of different modalities is presented and evaluated on three publicly real-world datasets with the task of cross-modal retrieval.
- –
- –
- –
- –
- –
- –
- An automatic and efficient pulmonary nodule detection system based on multi-model ensemble–
A novel full convolution segmentation framework for lung cavity extraction in preprocessing stage to solve the time consumption problem of the existing pulmonary nodule detection systems and a false positive reduction method based on multi-model ensemble is proposed for the further classification of nodule candidates.
- –
- Two-Headed Monster and Crossed Co-Attention Networks–
This paper proposes a paradigm, termed Two-Headed Monster (THM), which consists of two symmetric encoder modules and one decoder module connected with co-attention and designs Crossed Co-Attention Networks (CCNs) based on the Transformer model.
- –
- –
- –
- –
- –
- Intelligent Intrusion Detection Approach for SCADA System Protection–
Results show that the proposed OCSVM (One Class Support Vector Machine) based intrusion detection approach can be effectively used to detect both known and unknown attacks.
- –
- Preliminary construction of three dimensional multimedia platform of cheiloplasty–
A three-dimensional multimedia platform of cheiloplasty with lower triangular flap technique was established and may provide a new means for cheilplasty designing and clinical teaching.
- Factorial Hidden Markov Model Recognition Method Based on Multi-Channel Information Fusion–
The fault recognition in the speed-up and speed-down process of rotating machinery was successfully completed and the experiment result shows that the proposed FHMM fault recognition method is very effective.
- Using social annotations for trend discovery in scientific publications–
This work proposes a trend discovery process that employs trend estimation and trend selection and ranking for analyzing the emerging trends shown in the social annotation profiles, and adopts topic modeling on the annotation content to decapsulate the multitude of impact created by the annotated work.
- –
- –
- –
- A new pitch synchronous time domain phoneme recognizer using component analysis and pitch clustering–
A new clustering technique is designed that gives better results than other well known clustering algorithms like k-means, and is implemented to model pitch synchronous frames with gaussian mixture models.
- A Software for the Data Acquisition System in Guyed Mast–
A software for the data acquisition system for guyed mast with regard to the multifunction data acquisition card, and PSD (position sensitive detector) sensors are introduced.
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.