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
- Fahiem BacchusSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works369 from public data
- LEARNING BAYESIAN BELIEF NETWORKS: AN APPROACH BASED ON THE MDL PRINCIPLE903
A new approach for learning Bayesian belief networks from raw data is presented, based on Rissanen's minimal description length (MDL) principle, which can learn unrestricted multiply‐connected belief networks and allows for trade off accuracy and complexity in the learned model.
- A Survey on Aspect-Based Sentiment Analysis: Tasks, Methods, and Challenges547
A new taxonomy for ABSA is provided which organizes existing studies from the axes of concerned sentiment elements, with an emphasis on recent advances of compound ABSA tasks.
- Transformation Networks for Target-Oriented Sentiment Classification510
A new model is proposed to overcome the drawbacks of attention mechanism and the obstacles that block CNN to perform well in this sentiment classification task, and employs a CNN layer to extract salient features from the transformed word representations originated from a bi-directional RNN layer.
- A Unified Model for Opinion Target Extraction and Target Sentiment Prediction360
This paper aims to solve the complete task of target-based sentiment analysis in an end-to-end fashion, and presents a novel unified model which applies a unified tagging scheme.
- Neural Rating Regression with Abstractive Tips Generation for Recommendation346
A deep learning based framework named NRT is proposed which can simultaneously predict precise ratings and generate abstractive tips with good linguistic quality simulating user experience and feelings.
- 345
- Exploiting BERT for End-to-End Aspect-based Sentiment Analysis335
This work builds a series of simple yet insightful neural baselines to deal with E2E-ABSA, and shows that even with a simple linear classification layer, the BERT-based architecture can outperform state-of-the-art works.
- Aspect Sentiment Quad Prediction as Paraphrase Generation314
This work introduces the Aspect Sentiment Quad Prediction (ASQP) task, aiming to jointly detect all sentiment elements in quads for a given opinionated sentence, and proposes a novel Paraphrase modeling paradigm to cast the ASQP task to a paraphrase generation process.
- Towards Generative Aspect-Based Sentiment Analysis306
This paper proposes to tackle various ABSA tasks in a unified generative framework with two types of paradigms, namely annotation-style and extraction-style modeling, to enable the training process by formulating each ABSA task as a text generation problem.
- Aspect Term Extraction with History Attention and Selective Transformation290
Experimental results over four benchmark datasets clearly demonstrate that the new framework for tackling ATE can outperform all state-of-the-art methods.
- Deep Multi-Task Learning for Aspect Term Extraction with Memory Interaction260
A novel LSTM-based deep multi-task learning framework for aspect term extraction from user review sentences designed for jointly handling the extraction tasks of aspects and opinions via memory interactions is proposed.
- Graph Transformer for Graph-to-Sequence Learning257
A new model, known as Graph Transformer, is proposed that uses explicit relation encoding and allows direct communication between two distant nodes and provides a more efficient way for global graph structure modeling.
- Using a generalized instance set for automatic text categorization237
This work proposes a new technique known as the generalized instance set (GIS) algorithm by unifying the strengths of k-NN and linear classifiers and adapting to characteristics of text categorization problems.
- Deep Recurrent Generative Decoder for Abstractive Text Summarization218
A new framework for abstractive text summarization based on a sequence-to-sequence oriented encoder-decoder model equipped with a deep recurrent generative decoder (DRGN) achieves improvements over the state-of-the-art methods.
- Automatic text categorization and its application to text retrieval206
This work develops an automatic text categorization approach and investigates its application to text retrieval, demonstrating the effectiveness of the approach and demonstrating that the retrieval performance using automatic categorization achieves the same retrieval quality as the performance using manual categorization.
- A multilevel approach to intelligent information filtering199
A filtering model is proposed that decomposes the overall task into subsystem functionalities and highlights the need for multiple adaptation techniques to cope with uncertainties.
- Context-based generic cross-lingual retrieval of documents and automated summaries: Research Articles181
PRAW—a PRivAcy model for the Web suggested in this paper—is aimed at hiding users' navigation tracks to prevent eavesdroppers from inferring their profiles but still allowing them to be identified.
- Evaluation challenges in large-scale document summarization176
A large-scale meta evaluation of eight evaluation measures for both single-document and multi-document summarizers is presented, showing the strengths and draw-backs of all evaluation methods and how they rank the different summarizers.
- Unified Conversational Recommendation Policy Learning via Graph-based Reinforcement Learning169
A dynamic weighted graph based RL method to learn a policy to select the action at each conversation turn, either asking an attribute or recommending items, which significantly outperforms state-of-the-art methods but also enhances the scalability and stability of CRS.
- A Comprehensive Survey on Relation Extraction: Recent Advances and New Frontiers162
This survey provides a comprehensive review of existing deep learning techniques for Relation extraction and proposes a new taxonomy to categorize existing works from three perspectives, i.e., text representation, context encoding, and triplet prediction.
- Abstractive Multi-Document Summarization via Phrase Selection and Merging161
This work proposes an abstraction-based multi-document summarization framework that can construct new sentences by exploring more fine-grained syntactic units than sentences, namely, noun/verb phrases, and achieves reasonably well results on manual linguistic quality evaluation.
- Abstractive Multi-Document Summarization via Phrase Selection and148
This work proposes an abstraction-based multi-document summarization framework that can construct new sentences by exploring more fine-grained syntactic units than sentences, namely, noun/verb phrases, and achieves reasonably well results on manual linguistic quality evaluation.
- News Sensitive Stock Trend Prediction147
This paper investigates the immediate impact of news articles on the time series based on the Efficient Markets Hypothesis and proposes an algorithm, which is called guided clustering, to filter news articles with the help of the clusters that have been obtained from trends.
- 141
- A Theoretical Analysis of the Repetition Problem in Text Generation130
A new framework for theoretical analysis for the repetition problem is proposed and a novel rebalanced encoding approach is proposed to alleviate the high inflow problem and thus reducing the upper bound.
- Jointly Identifying Entities and Extracting Relations in Encyclopedia Text via A Graphical Model Approach130
A joint discriminative probabilistic model with arbitrary graphical structure to optimize all relevant subtasks simultaneously and a new inference method, namely collective iterative classification (CIC), to find the most likely assignments for both entities and relations are proposed.
- 126
- Daily prediction of Major Stock Indices from Textual WWW Data126
The authors predict stock markets using information contained in articles published on the Web, mostly textual articles appearing in the leading and the most influential financial newspapers, to produce daily closing values of major stock market indices in Asia, Europe and America.
- AMR Parsing via Graph-Sequence Iterative Inference124
A new end-to-end model that treats AMR parsing as a series of dual decisions on the input sequence and the incrementally constructed graph based on iterative inference that helps achieve better answers in both perspectives, leading to greatly improved parsing accuracy.
- Neural Machine Translation with Monolingual Translation Memory121
A new framework that uses monolingual memory and performs learnable memory retrieval in a cross-lingual manner and outperforms strong TM-augmented NMT baselines using bilingual TM.
- Using evolutionary programming and minimum description length principle for data mining of Bayesian networks119
A new approach to learning Bayesian network structures based on the minimum description length (MDL) principle and evolutionary programming is developed, which employs a MDL metric and integrates a knowledge-guided genetic operator for the optimization in the search process.
- VisualWebBench: How Far Have Multimodal LLMs Evolved in Web Page Understanding and Grounding?110
This paper introduces \bench{}, a multimodal benchmark designed to assess the capabilities of MLLMs across a variety of web tasks, and evaluates 14 open-source MLLMs, Gemini Pro, Claude-3 series, and GPT-4V(ision) on \bench{}, revealing significant challenges and performance gaps.
- CodeAttack: Revealing Safety Generalization Challenges of Large Language Models via Code Completion104
CodeAttack is introduced, a framework that transforms natural language inputs into code inputs, presenting a novel environment for testing the safety generalization of LLMs and highlighting new safety risks in the code domain and the need for more robust safety alignment algorithms to match the code capabilities of LLMs.
- Skeleton-to-Response: Dialogue Generation Guided by Retrieval Memory99
This paper proposes a new framework which exploits retrieval results via a skeleton-to-response paradigm and shows that this approach significantly improves the informativeness of the generated responses.
- Detection of shifts in user interests for personalized information filtering98
Detection of Shifts in User Interests for Personalized Information Filtering W. Lam*, S. Mukhopadhyay, J. Mostafa, and M. Palakal .
- Using Causal Information and Local Measures to Learn Bayesian Networks92
A new local way of computing the description length is presented, which allows for significant improvements in the search algorithm and opens the door for local refinement of an existent network.
- Proactive Conversational AI: A Comprehensive Survey of Advancements and Opportunities90
A comprehensive review of various prominent problems and advanced designs for implementing proactivity into different types of dialogue systems, including open-domain dialogues, task-oriented dialogues, and information-seeking dialogues are provided.
- Discovering useful concept prototypes for classification based on filtering and abstraction90
This work develops anew framework, called ICPL (Integrated Concept Prototype Learner), which integrates instance-filtering and instance-abstraction techniques by maintaining a balance of different kinds of concept prototypes according to instance locality.
- Jointly Learning Word Embeddings and Latent Topics83
The experimental results demonstrate that the STE model can indeed generate useful topic-specific word embeddings and coherent latent topics in an effective and efficient way.
- Detection and prevalence of the “f” variant of epstein-barr virus in Southern China82
The "f" variant of Epstein-Barr virus may have an association with the development and/or maintenance of nasopharyngeal carcinoma (NPC) among Southern Chinese and this variant is detected at a higher frequency among individuals with elevated IgA antibody levels against EBV capsid antigen and in NPC patients as compared to healthy individuals or patients who are in remission for NPC for over 3 years.
- Cross-lingual Aspect-based Sentiment Analysis with Aspect Term Code-Switching81
This paper considers the unsupervised cross-lingual transfer for the ABSA task, where only labeled data in the source language is available and aims at transferring its knowledge to the target language having no labeled data and proposes an alignment-free label projection method.
- Phoneme-Based Transliteration of Foreign Names for OOV Problem81
This work presents a statistical transliteration method for automatic translation according to pronunciation similarities, i.e. to map phonemes comprising an English name to the phonetic representations of the corresponding Chinese name.
- A Unified Multi-task Learning Framework for Multi-goal Conversational Recommender Systems80
This work proposes a novel Unified MultI-goal conversational recommeNDer system (UniMIND), which unify these four tasks with different formulations into the same sequence-to-sequence paradigm and achieves state of the art performance on all tasks with a unified model.
- Extracting discriminative concepts for domain adaptation in text mining75
This work proposes a domain adaptation method that parameterizes this concept space by linear transformation under which it explicitly minimize the distribution difference between the source domain with sufficient labeled data and target domains with only unlabeled data, while at the same time minimizing the empirical loss on the labeled data in the sourcedomain.
- A new hybrid method for learning bayesian networks: Separation and reunion72
This paper proposes a new hybrid Bayesian network learning method, SAR (the abbreviation of Separation And Reunion), which maintains the merits of both types of learning techniques while avoiding their drawbacks.
- Decoder-Only or Encoder-Decoder? Interpreting Language Model as a Regularized Encoder-Decoder69
This paper proposes a novel partial attention language model to solve the attention degeneration problem in the decoder-only language model framework through the analysis of a regularized encoder-decoder structure that has an encoder and a decoder making it easier to be compared with the classical Encoder-Decoder structure.
- Using New Data to Refine a Bayesian Network**Wai Lam's work was supported by an OGS scholarship. Fahiem Bacchus's work was supported by the Canadian Government through the NSERC and IRIS programs. The authors' e-mail addresses are [wlaml|fbacchus]@logos.uwaterloo.ca.69
This work develops a new approach for refining the network's structure based on the Minimal Description Length (MDL) principle, and it employs an adapted version of a Bayesian network learning algorithm developed in the previous work.
- CoreGen: Contextualized Code Representation Learning for Commit Message Generation68
CoreGen first learns contextualized code representation which exploits the contextual information behind code commit sequences and the learned representations of code commits built upon Transformer are then transferred for downstream commit message generation.
- Contextualized Code Representation Learning for Commit Message Generation68
CoreGen first learns contextualized code representation which exploits the contextual information behind code commit sequences and the learned representations of code commits built upon Transformer are then transferred for downstream commit message generation.
- An unsupervised framework for extracting and normalizing product attributes from multiple web sites67
An unsupervised framework for simultaneously extracting and normalizing attributes of products from multiple Web pages originated from different sites based on a probabilistic graphical model based on variational method, which can be applied to a wide range of Web mining applications such as product matching and retrieval.
- Joint Learning of Answer Selection and Answer Summary Generation in Community Question Answering65
A question-driven pointer-generator network is designed, which exploits the correlation information between question-answer pairs to aid in attending the essential information when generating answer summaries and a new large-scale CQA corpus is constructed, WikiHowQA, which contains long answers for answer selection as well as reference summaries for answer summarization.
- Meta-evaluation of summaries in a cross-lingual environment using content-based metrics65
A framework for the evaluation of summaries in English and Chinese using similarity measures that can be used to evaluate extractive, non-extractive, single and multi-document summarization is described.
- Medical data mining using evolutionary computation63
A system for discovering medical knowledge by learning Bayesian networks and rules is introduced and is applied to real-life medical databases for limb fracture and scoliosis.
- Salience Estimation via Variational Auto-Encoders for Multi-Document Summarization62
An unsupervised data reconstruction framework is proposed, which jointly considers the reconstruction for latent semantic space and observed term vector space and can capture the salience of sentences from these two different and complementary vector spaces.
- Bayesian network refinement via machine learning approach61
An approach to refining Bayesian network structures from new data by a machine learning approach based on a formalism known as the minimum description length (MDL) principle, which has the capability of refining a network structure using partially specified data.
- Conversational Fashion Image Retrieval via Multiturn Natural Language Feedback60
A novel framework that can effectively handle conversational fashion image retrieval with multiturn natural language feedback texts is proposed that searches for candidate images based on exploitation of the encoded reference image and feedback text information together with the conversation history.
- 60
- 60
- An unsupervised topic segmentation model incorporating word order58
This work presents a new unsupervised topic discovery model for a collection of text documents that does not break the document's structure such as paragraphs and sentences and preserves word order, and can generate two levels of topics of different granularity.
- An active learning framework for semi-supervised document clustering with language modeling58
A gain-directed document pair selection method that measures how much the authors can learn by revealing judgments of selected document pairs is designed and uses the estimation of term co-occurrence probabilities as a clue for finding informative document pairs.
- Core Semantic First: A Top-down Approach for AMR Parsing56
A novel scheme for parsing a piece of text into its Abstract Meaning Representation (AMR): Graph Spanning based Parsing (GSP), which achieves the state-of-the-art performance in the sense that no heuristic graph re-categorization is adopted.
- Automatic classification using supervised learning in a medical document filtering application56
The effectiveness of a classifier that uses supervised learning was analyzed in terms of its accuracy and ultimately its influence on filtering in order to clarify the impact of this performance on filtering.
- A meta-learning approach for text categorization54
The results confirm that the meta-model approach can exploit the advantage of its component algorithms, and demonstrate a better performance than existing algorithms.
- Chain-of-Symbol Prompting Elicits Planning in Large Langauge Models51
A novel method called C O S is proposed that represents the complex environments with condensed symbolic spatial representations during the chained intermediate thinking steps that surpasses the performance of the Chain-of-Thought Prompting in all three planning tasks with even fewer tokens used in the inputs.
- Using contextual analysis for news event detection51
The results show that the performance of detection using concept terms together with story keywords is better than traditional methods which only use keyword representation.
- Developing Infrastructure for the Evaluation of Single and Multi-document Summarization Systems in a Cross-lingual Environment50
This work describes the development of Language and Evaluation Resources for the evaluation of summaries in English and Chinese and focuses on the resources developed that are made available for the research community.
- Toward Personalized Answer Generation in E-Commerce via Multi-perspective Preference Modeling49
Experimental results on real-world E-commerce QA datasets demonstrate that the proposed Personalized Answer GEneration method outperforms existing methods by generating informative and customized answers and show that answer generation in E-Commerce can benefit from personalization.
- 46
- Web Query Reformulation via Joint Modeling of Latent Topic Dependency and Term Context45
This article proposes a graphical model to score queries that exploits a latent topic space, which is automatically derived from the query log, to detect semantic dependency of terms in a query and dependency among topics.
- Learning to extract and summarize hot item features from multiple auction web sites45
A unified framework which can automatically extract product features and summarize hot item features from multiple auction sites is developed, which can model the inter-dependence between neighbouring tokens in a Web page, tokens in different Web pages, as well as various information such as hot Item features across different auction sites.
- 45
- Unsupervised Extraction of Popular Product Attributes from E-Commerce Web Sites by Considering Customer Reviews44
An unsupervised learning framework for extracting popular product attributes from product description pages originated from different E-commerce Web sites that is able to not only detect popular product features from a collection of customer reviews but also map these popular features to the related product attributes.
- 43
- Cascaded Attention based Unsupervised Information Distillation for Compressive Summarization43
A cascaded attention based unsupervised model to estimate the salience information from the text for compressive multi-document summarization and achieves better results than the state-of-the-art methods.
- An unsupervised method for joint information extraction and feature mining across different Web sites43
An unsupervised learning framework which can jointly extract information and conduct feature mining from a set of Web pages across different sites, based on an undirected graphical model which can model the interdependence between the text fragments within the same Web page, as well as text fragments in different Web pages.
- Tackling Long-Tailed Relations and Uncommon Entities in Knowledge Graph Completion42
A meta-learning framework that aims at handling infrequent relations with few-shot learning and uncommon entities by using textual descriptions is proposed and a novel model is designed to better extract key information from textual descriptions.
- Extracting causation knowledge from natural language texts42
This work investigates the application of SEKE on two domain areas of news articles, namely the Hong Kong stock market movement domain and the global warming domain, and finds that SEKE can effectively handle the natural language texts in these two domains for causation knowledge extraction.
- 41
- Dynamic Semantic Graph Construction and Reasoning for Explainable Multi-hop Science Question Answering41
Results on two scientific multi-hop QA datasets show that the proposed framework can surpass recent approaches including those using additional knowledge graphs while maintaining high explainability on OpenBookQA and achieve a new state-of-the-art result on ARC-Challenge in a computationally practicable setting.
- 41
- User Satisfaction Estimation with Sequential Dialogue Act Modeling in Goal-oriented Conversational Systems39
This paper proposes a novel framework, namely USDA, to incorporate the sequential dynamics of dialogue acts for predicting user satisfaction, by jointly learning User Satisfaction Estimation and Dialogue Act Recognition tasks.
- Review-Aware Answer Prediction for Product-Related Questions Incorporating Aspects39
A new framework that can tackle the task of review-aware answer prediction for product-related questions and is able to predict the answers for new yes-no questions taking into consideration of aspects is proposed.
- 39
- 38
- Collaborative Filtering Incorporating Review Text and Co-clusters of Hidden User Communities and Item Groups38
A new generative model is proposed to predict user's ratings on previously unrated items by considering review texts as well as hidden user communities and item groups relationship, and co-clustering would be more effective for modeling the generation of review texts.
- Joint Modeling of Participant Influence and Latent Topics for Recommendation in Event-based Social Networks37
An event recommendation model is proposed that considers participant influence, and exploits the influence of existing participants on the decisions of new participants based on Poisson factorization, which achieves superior event recommendation performance over several state-of-the-art models.
- Neuro-Symbolic Integration Brings Causal and Reliable Reasoning Proofs36
A neuro-symbolic integration method is presented, in which a neural LLM is used to represent the knowledge of the problem while an LLM-free symbolic solver is adopted to do deliberative reasoning using the knowledge.
- A distributed learning algorithm for Bayesian inference networks36
This work presents a new distributed algorithm for computing the minimum description length (MDL) in learning Bayesian inference networks from data, which exploits both properties of the MDL-based score metric and a distributed, asynchronous, adaptive search technique called nagging.
- Answer Ranking for Product-Related Questions via Multiple Semantic Relations Modeling35
This paper proposes an answer ranking model named MUSE which carefully models multiple semantic relations among the question, answers, and relevant reviews, and achieves superior performance on the concerned answer ranking task.
- 35
- Learning good prototypes for classification using filtering and abstraction of instances35
It is found that the PGF framework maintains or achieves better classification accuracy and gains a significant improvement in data reduction compared with pure filtering and pure abstraction techniques as well as KNN and C4.5.
- Nonfactoid Question Answering as Query-Focused Summarization With Graph-Enhanced Multihop Inference34
A novel query-focused summarization method, namely a graph-enhanced multihop query- focused summarizer (GMQS), to tackle the nonfactoid QA problem and possesses the capability of performing explainable multihip reasoning.
- QuaSE: Sequence Editing under Quantifiable Guidance34
This framework explores the pseudo-parallel sentences by modeling their content similarity and outcome differences to enable a better disentanglement of the latent factors, which allows generating an output to better satisfy the desired outcome and keep the content.
- Towards a unified solution34
This paper proposes a unified method to tackle the task of data record extraction from Web pages by addressing several key issues in a uniform manner and achieves higher accuracy compared with three state-of-the-art methods.
- 34
- Opinion-aware Answer Generation for Review-driven Question Answering in E-Commerce33
Two kinds of opinion fusion strategies, namely, static and dynamic fusion, are proposed to distill and aggregate important opinion information learned from the opinion mining task into the answer generation process.
- 33
- Exploiting Reasoning Chains for Multi-hop Science Question Answering32
The proposed Chain Guided Retriever-reader framework allows the retriever to capture step-by-step clues of the entire reasoning process, which is not only shown to be effective on two challenging multi-hop Science QA tasks, namely OpenBookQA and ARC-Challenge, but also favors explainability.
- AnswerFact: Fact Checking in Product Question Answering32
This paper introduces AnswerFact, a large scale fact checking dataset from product question answering forums, and proposes a novel neural model with tailored evidence ranking components to handle the concerned answer veracity prediction problem.
- Wikipedia entity expansion and attribute extraction from the web using semi-supervised learning32
A semi-supervised learning model with Conditional Random Fields is developed to deal with the issues of extraction learning and limited number of labeled examples derived from the seed entities to conduct discovery and extraction from desirable semi-structured data record sets.
- 32
- 32
- Harnessing Webpage UIs for Text-Rich Visual Understanding31
This work proposes synthesizing general multimodal instructions from webpage UIs using text-based large language models, a dataset containing 7.3 million samples from 1 million websites, covering diverse multimodal tasks and UI layouts that generalize surprisingly well to non-web UI tasks and even to non-UI domains, such as document understanding, OCR, and chart interpretation.
- 31
- Learning Domain-Sensitive and Sentiment-Aware Word Embeddings31
Experimental results show that the proposed model provides an effective way to learn domain-sensitive and sentiment-aware word embeddings which benefit sentiment classification at both sentence level and lexicon term level.
- Aspect-based Sentiment Analysis in Question Answering Forums30
A model with a design of cross-sentence aspect-opinion interaction modeling to address the task of ABSA on QA forums (ABSA-QA), aiming to jointly detect the discussed aspects and their sentiment polarities for a given QA pair.
- Using query log and social tagging to refine queries based on latent topics30
This paper proposes a graphical model to score queries that exploits a latent topic space, which is automatically derived from the query log, and makes it feasible to score some queries which do not have much available historical term context information.
- Learning nonlinear multiregression networks based on evolutionary computation29
The empirical results demonstrate that the novel knowledge discovery and data mining framework can successfully discover the target network structure and the regression coefficients, and propose a global learning algorithm, called MRCEP-FDOA, for discovering the network structures and their elements from databases.
- Contextualized Knowledge-aware Attentive Neural Network: Enhancing Answer Selection with Knowledge28
A context-knowledge interaction learning framework, Knowledge-aware Neural Network, which learns the QA sentence representations by considering a tight interaction with the external knowledge from KG and the textual information and develops two kinds of knowledge-aware attention mechanism to summarize both the context-based and knowledge-based interactions between questions and answers.
- Using Support Vector Machines for Terrorism Information Extraction28
This paper proposes to model the content and context information of a candidate entity to be extracted as a set of features and builds a classification model for each category of entities using Support Vector Machines (SVM).
- A fuzzy expert database system27
A fuzzy expert database system which is an integration of a fuzzy expert system building tool called SYSTEM Z-II and a database management system called Rdb/VMS, able to extract fuzzy data and terms stored in a database and used in the fuzzy reasoning in an expert system.
- 26
- Evaluation of Text Summarization in a Cross-lingual Information Retrieval Framework26
A toolkit for evaluation of single-document and multi-document summarization and evaluation of summarization in the framework of cross-lingual information retrieval is developed and the measurement of relevance correlation is introduced and systematically examined in this workshop.
- 25
- Reader-Aware Multi-Document Summarization: An Enhanced Model and The First Dataset25
Experimental results show that reader comments can improve the summarization performance, which also demonstrates the usefulness of the proposed dataset.
- FIDS: an intelligent financial Web news articles digest system25
A system called FIDS (Financial Information Digest System), which can digest online financial news automatically and allows one to perform cross-validation on their contents, so users can have access to more complete information which otherwise would be scattered in different articles.
- Asking Multimodal Clarifying Questions in Mixed-Initiative Conversational Search24
It is hypothesized that in scenarios where multimodal information is pertinent, the clarification process can be improved by using non-textual information, and it is proposed to add images to clarifying questions and formulate the novel task of asking multimodal clarifying questions in open-domain, mixed-initiative conversational search systems.
- Partially-Aligned Data-to-Text Generation with Distant Supervision24
A novel distant supervision generation framework that firstly estimates the input data's supportiveness for each target word with an estimator and then applies a supportiveness adaptor and a rebalanced beam search to harness the over-generation problem in the training and generation phases respectively.
- 24
- Adapting Web information extraction knowledge via mining site-invariant and site-dependent features24
A novel framework that aims at automatically adapting previously learned information extraction knowledge from a source Web site to a new unseen target site in the same domain and derives site-invariant features from previously learned extraction knowledge and the items previously collected or extracted from the source web site is developed.
- Learning phonetic similarity for matching named entity translations and mining new translations24
A novel named entity matching model which considers both semantic and phonetic clues is proposed and a mining framework for discovering new, unseen named entity translations from online daily Web news is developed.
- Efficient in-memory extensible inverted file23
An analytical investigation and an empirical evaluation of storage-efficient inmemory extensible inverted files, which are represented by fixed- or variable-sized linked list nodes, and found that the arrival rate scheme is the best in terms of both storage utilization and the mean number of chainings per term.
- Direct identification of serotypes of natural human rotavirus isolates by hybridization using cDNA probes derived from segment 9 of the rotavirus genome23
Under stringent hybridization conditions, cDNA of segment 9 of the rotavirus genome, which codes for the viral protein VP7, permitted differentiation of serotypes of culture-grown rotaviruses and natural isolates of human rotviruses directly from clinical specimens.
- Reasons to Reject? Aligning Language Models with Judgments22
A novel framework, Contrastive Unlikelihood Training (CUT), is proposed that allows for fine-grained inappropriate content detection and correction based on judgments, and suggests that judgments hold greater potential than rewards in LLM alignment.
- Multi-hop Inference for Question-driven Summarization22
This work proposes a novel question-driven abstractive summarization method, Multi-hop Selective Generator (MSG), to incorporate multi-hop reasoning into question- driven summarization and, meanwhile, provide justifications for the generated summaries.
- Review-guided Helpful Answer Identification in E-commerce22
The Review-guided Answer Helpfulness Prediction (RAHP) model is proposed that not only considers the interactions between QA pairs but also investigates the opinion coherence between the answer and crowds’ opinions reflected in the reviews, which is another important factor to identify helpful answers.
- Detecting Common Discussion Topics Across Culture From News Reader Comments22
A new probabilistic graphical model called MCTA is proposed which can cope with the language gap and capture the common semantics in different languages and a partially collapsed Gibbs sampler is developed which effectively incorporates the term translation relationship into the detection of cultural-common topics for model parameter learning.
- A unified model for unsupervised opinion spamming detection incorporating text generality22
This work proposes a unified probabilistic graphical model to detect the suspicious review spams, the review spammers and the manipulated offerings in an unsupervised manner and demonstrates the superiority of the proposed model compared with the state-of-the-art models.
- 22
- Partner Personas Generation for Dialogue Response Generation21
Experimental results from automatic and human evaluations indicate that this framework is capable of generating relevant, interesting, coherent and informative partner personas, even compared to the ground truth partner persona, which enhances the succeeding dialogue response generation.
- Multilingual AMR Parsing with Noisy Knowledge Distillation21
This work obtains an AMR parser whose performances surpass all previously published results on four different foreign languages, including German, Spanish, Italian, and Chinese, by large margins.
- Information Retrieval Technology21
This book constitutes the refereed proceedings of the 15th Information Retrieval Technology Conference, AIRS 2019, held in Hong Kong, China, in November 2019.
- 21
- Once Upon a $\textit{Time}$ in $\textit{Graph}$: Relative-Time Pretraining for Complex Temporal Reasoning20
This work proposes RemeMo, which explicitly connects all temporally-scoped facts by modeling the time relations between any two sentences, and suggests creating a graph structure based on the relative placements of events along the time axis.
- Leveraging Long Short-Term User Preference in Conversational Recommendation via Multi-agent Reinforcement Learning20
This work proposes two novel multi-agent RL-based frameworks, namely Independent and Hierarchical Multi-Agent UNIfied COnversational RecommeNders (IMA-UNicORN and HMA-UNICORN), respectively, which significantly outperform state-of-the-art methods on the multi-round conversational recommendation.
- Review-based Question Generation with Adaptive Instance Transfer and Augmentation20
This work proposes to explore question generation as a new way of review information exploitation, namely generating questions that can be answered by the corresponding review sentences, with an iterative learning framework with adaptive instance transfer and augmentation.
- 20
- Learning to extract hierarchical information from semi-structured documents20
This work develops a new method, known as HISER (HIerarchical record Structure and Extraction Rule learning), which employs a tw ostage learning task, namely, hierarc hical record structure learning and extraction rule learning.
- Adaptive Concept Resolution for document representation and its applications in text mining19
This work presents an Adaptive Concept Resolution (ACR) model, which can learn a concept border from an ontology taking into the consideration of the characteristics of the particular document collection, and outperforms an existing static method in almost all cases.
- Direct detection of Epstein‐Barr virus in peripheral blood and comparison of Epstein‐Barr virus genotypes present in direct specimens and lymphoblastoid cell lines established from nasopharyngeal carcinoma patients and healthy carriers in Hong Kong19
This finding suggests selective isolation of prototype F EBV in spontaneous LCLs established from NPC patients' peripheral blood, but was only detected in 2/9 specimens directly analyzed by the PCR assay.
- Contrastive Representation Learning for Exemplar-Guided Paraphrase Generation18
A new method to design two contrastive losses with respect to the content and the style by considering two problem characteristics during training with the goal of learning a better representation of the style and the content.
- Bridging Hierarchical and Sequential Context Modeling for Question-driven Extractive Answer Summarization18
This paper designs a hierarchical compare-aggregate method to integrate the interaction between QA pairs in both word-level and sentence-level into the final question and answer representations and conducts the question-aware sequential extractor to produce a summary for the lengthy answer.
- Supervised topic models with word order structure for document classification and retrieval learning18
A novel supervised topic model for document retrieval learning is proposed which can be regarded as a pointwise model for tackling the learning-to-rank task and improves upon the state-of-the-art models.
- Web page segmentation with structured prediction and its application in web page classification18
A learning algorithm based on the structured output Support Vector Machine framework is developed to determine the feature weights, which is capable to consider the inter-dependency among candidate segmentation boundaries.
- Chinese NER Using CRFs and Logic for the Fourth SIGHAN Bakeoff18
A high-performance Chinese NER system that incorporates Conditional Random Fields and first-order logic for the fourth SIGHAN Chinese language processing bakeoff (SIGHAN-6) is reported, achieving consistently high performance and showing both the attractiveness and effectiveness of the proposed model.
- Named entity translation matching and learning18
A mining system is developed for discovering new named entity translations from daily Web news that cannot be found in the existing bilingual dictionary using a named entity matching model that makes use of both semantic and phonetic evidence.
- Improving Transliteration with Precise Alignment of Phoneme Chunks and Using Contextual Features18
This paper presents a direct probabilistic transliteration model using contextual features of phonemes with a tailored alignment scheme for phoneme chunks with superior performance over the source-channel for the task of English-Chinese transliterations.
- A Frustratingly Simple Decoding Method for Neural Text Generation17
Frustratingly Simple Decoding is introduced, a frustratingly simple, highly efficient, and surprisingly effective decoding method for neural text generation that outperforms established strong baselines in terms of generation quality, decoding speed, and universality.
- Relevance-Promoting Language Model for Short-Text Conversation17
A relevance-promoting transformer language model is designed, which performs additional supervised source attention after the self-attention to increase the importance of informative query tokens in calculating the token-level representation.
- 17
- 16
- Learning Belief Networks for Language Understanding16
Experiments with the ATIS-3 corpus shows that the enhanced topology improves goal identification accuracies from 83.7% to 91.5% when a single output goal is evaluated, and from 66.0% to 83.1% when multiple output goals are evaluated.
- Responding E-commerce Product Questions via Exploiting QA Collections and Reviews15
A new framework for automatically responding product questions newly posed by users via exploiting existing QA collections and review collections in a coordinated manner is proposed, which can return a ranked list of snippets serving as the automated response for a given question.
- Normalizing web product attributes and discovering domain ontology with minimal effort15
A framework aiming at normalizing product attributes from Web pages collected from different Web sites without the need of labeled training examples and a method to automatically construct a domain ontology using the normalized product attributes which are the output of the inference on the graphical model.
- 14
- 14
- Mining employment market via text block detection and adaptive cross-domain information extraction14
The approach is able to extract precise information from the text content supporting useful employment market analysis locally and globally and can easily adapt to a large number of online job advertisements in different and new domains.
- Analysis of Epstein—Barr virus in localized nasopharyngeal carcinoma tumors14
An f variant of Epstein‐Barr virus appears associated with nasopharyngeal carcinoma (NPC) in Southern Chinese and early diagnosis of the tumor allows the detection of some localized tumors.
- Exploring Compositional Generalization of Large Language Models13
It is argued that the ability to generalize from simple instructions to more intricate compositional instructions represents a key aspect of the out-of-distribution generalization for LLMs.
- Learning a unified embedding space of web search from large-scale query log13
This paper develops the well-known Word2vec to perform the learning of the semantic space with search session data, and uses a set of session graphs to represent search sessions, and the space learning task is cast as a vector learning problem for the graph vertices by maximizing the log-likelihood of a training session data set.
- Latent Aspect Mining via Exploring Sparsity and Intrinsic Information13
A new generative model is proposed to tackle the latent aspect mining problem in an unsupervised manner and introduces two latent variables, namely, user intrinsic aspect interest and item intrinsic aspect quality facilitating better modeling of aspect generation leading to improvement on the accuracy and reliability of predicted aspect ratings.
- Unsupervised Extraction of Popular Product Attributes from Web Sites13
A framework which is able not only to detect concerned popular features of a product from a collection of customer reviews, but also to map these popular features to the related product attributes, and at the same time to extract these attributes from description pages is proposed.
- Using Semantic Relations with World Knowledge for Question Answering.13
This work investigates the possibilities of applying linguistically sophisticated grammatical framework in tackling the real-world natural language processing task such as question answering by exploiting the possible world’s entities and relations as described in online encyclopedia.
- Data Augmentation Based on Adversarial Autoencoder Handling Imbalance for Learning to Rank12
A data generation model based on Adversarial Autoencoder (AAE) for tackling the data imbalance in LTR via informative data augmentation via adversarial training stage and can significantly improve the performance of LTR models.
- 12
- Sentiment Analysis of Fashion Related Posts in Social Media11
This work proposes a novel framework that jointly leverages the image vision, post text, as well as fashion attribute modality to determine the sentiment category and exploits the mutual relationship between the fashion attributes and the post texts via a mutual attention mechanism.
- Open Domain Event Text Generation11
A novel random drop component is proposed that randomly deletes words from the retrieved sentences making the model more robust for handling long input sentences and the experimental results show that the carefully-designed architecture does help generate better event text.
- Who Wants to Join Me?: Companion Recommendation in Location Based Social Networks11
This model is the first that addresses the task of recommending companions for a proposed activity and uses a latent topic model, which moreover takes into account the geographic proximity of the user to the location of the proposed venue.
- Web Entity Detection for Semi-structured Text Data Records with Unlabeled Data11
A framework for named entity detection from Web content associated with semi-structured text data records is proposed, by exploiting the inherent structure via a transformation process facilitating collective detection.
- Ontology enhancement and concept granularity learning11
A method to enhance WordNet automatically by merging Wikipedia entities into WordNet, and construct an enriched ontology, named as WorkiNet, which keeps the desirable structure of WordNet and generates a tailor-made semantic concept collection for a given document collection.
- 11
- 11
- 10
- 10
- Robust detection of semi-structured web records using a DOM structure-knowledge-driven model10
This work proposes a Structure-Knowledge-Oriented Global Analysis (Skoga) framework which can perform robust detection of different-kinds of data records and record regions and achieves higher accuracy compared with state-of-the-art methods.
- An integrated discriminative probabilistic approach to information extraction10
This integrated model has several advantages over previous probabilistic graphical models, and it offers a great capability of extracting implicit relations and new relation discovery for relation extraction from encyclopedic documents, and capturing sub-structures in named entities for named entity recognition.
- An integrated probabilistic and logic approach to encyclopedia relation extraction with multiple features10
A new integrated approach based on Markov logic networks (MLNs), an effective combination of probabilistic graphical models and first-order logic for statistical relational learning, to extracting relations between entities in encyclopedic articles from Wikipedia significantly outperforms the current state-of-the-art probabilists.
- Hidden dynamic probabilistic models for labeling sequence data10
A new discriminative framework, namely Hidden Dynamic Conditional Random Fields (HDCRFs), for building probabilistic models which can capture both internal and external class dynamics to label sequence data is proposed.
- Customized classification learning based on query projections10
A customized classification learning method QPL, which is based on query projections, which explores the projections of the query instance (QPs), which are essentially subsets of attribute values shared by the query and training instances.
- 10
- Extracting Causation Knowledge from Natural Language Texts10
To demonstrate the flexibility of SEKE2 for different domains, this work studies the application of causation semantic templates on two domain areas of news stories, namely, Hong Kong stock market movement and global warming.
- 9
- Entity Retrieval in the Knowledge Graph with Hierarchical Entity Type and Content9
This model directly encodes entity types and entity descriptions into a Markov random field based framework via a path aware smoothing method and achieves improvements over the existing and state-of-the-art models.
- A Framework Based on Graphical Models with Logic for Chinese Named Entity Recognition9
Experimental results show that the proposed framework of probabilistic graphical models with first-order logic significantly outperforms the state-of-the-art models for solving this task.
- 9
- 9
- From Cloze to Comprehension: Retrofitting Pre-trained Masked Language Model to Pre-trained Machine Reader8
Although conceptually simple, PMR is particularly effective in solving extraction tasks including Extractive Question Answering and Named Entity Recognition, where it shows tremendous improvements over previous approaches especially under low-resource settings.
- 8
- On Controlling Fallback Responses for Grounded Dialogue Generation8
A novel framework that automatically generates a control token with the generator to bias the succeeding response towards informativeness for answerable contexts and fallback for unanswerable contexts in an end-to-end manner is proposed.
- Product Question Intent Detection using Indicative Clause Attention and Adversarial Learning8
This work conducts a thorough investigation on product questions covering different product categories from some commercial E-commerce sites and develops an intent detection framework based on a tailor-made deep neural model capable of coping with long and verbose questions via identifying the indicative clauses.
- Entity Retrieval via Entity Factoid Hierarchy8
This work proposes a new entity representation model named as entity factoid hierarchy, which is a tree structure composed of factoid nodes that describes one or more facts about the entity in different information granularities.
- An N-Gram Topic Model for Time-Stamped Documents8
A topic model that captures the temporal dynamics in the text data along with topical phrases and derives a collapsed Gibbs sampler for posterior inference to show improvement over the current state-of-the-art topics over time model.
- Probabilistic joint models incorporating logic and learning via structured variational approximation for information extraction8
A joint discriminative probabilistic framework to optimize all relevant subtasks simultaneously and provides a great flexibility to incorporate the advantage of both uncertainty for sequence modeling and first-order logic for domain knowledge.
- 8
- 8
- Context‐based generic cross‐lingual retrieval of documents and automated summaries8
A context-based generic cross-lingual retrieval model that can deal with different language pairs based on co-occurrence statistics from different granularity of passages is developed and implemented.
- APEX-SQL: Talking to the data via Agentic Exploration for Text-to-SQL7
APEX-SQL is proposed, an Agentic Text-to-SQL Framework that shifts the paradigm from passive translation to agentic exploration, and introduces a deterministic mechanism to retrieve exploration directives, allowing the agent to effectively explore data distributions, refine hypotheses, and generate semantically accurate SQLs.
- Bridging the Gap between Pre-Training and Fine-Tuning for Commonsense Generation7
This paper designs a new format of input to endow PLMs the ability to deal with masked sentences with incorrect word order and inserts the special token [MASK] between two consecutive concept words to make theinput distribution more similar to the input distribution in pre-training.
- 7
- Exploiting interactions of review text, hidden user communities and item groups, and time for collaborative filtering7
A new model is developed to tackle the CF problem which predicts user’s ratings on previously unrated items by effectively exploiting interactions among review texts as well as the hidden user community and item group information and proposes an enhanced model called TCMR to consider time information and exploit the temporal interactions between review texts and co-clusters of user communities and item groups.
- Web Discovery and Filtering Based on Textual Relevance Feedback Learning7
WID allows users to provide three forms of the relevance feedback model, namely, the positive page feedback, the negativepage feedback, and the positive keyword feedback, which is domain independent and does not rely on any prior knowledge or information about the Web content.
- Automatic Textual Document Categorization Using Multiple Similarity-Based Models7
A meta-model framework which combines the strength of different variants of GIS algorithm as well as state-of-the-art existing algorithms using multivariate regression analysis on document feature characteristics is proposed.
- Prototype Generation Based on Instance Filtering and Averaging7
A new algorithm, called Prototype Generation and Filtering (PGF), which combines the strength of instance-filtering and instance-averaging techniques, which is able to generate representative prototypes while eliminating noise and exceptions is proposed.
- Identification of technically challenging variants: Whole-genome sequencing improves diagnostic yield in patients with high clinical suspicion of rare diseases6
The study demonstrated the indispensable diagnostic power of WGS in detecting technically challenging variants and the capability to serve as an all-in-one test for patients with high clinical suspicion of rare diseases.
- 6
- Joint learning of text alignment and abstractive summarization for long documents via unbalanced optimal transport6
A novel NATS framework built based on the theory of unbalanced optimal transport (UOT), named as UOTSumm, which directly learns the text alignment from summarization data, without resorting to any biased tool such as ROUGE.
- Learning to Ask Clarification Questions with Spatial Reasoning6
A novel method is proposed, named SpatialCQ, to align the representation space between textual and spatial information by encoding spatial states with textual descriptions and fuse the multimodal information for asking clarification questions.
- 6
- 6
- 6
- Learning to Rank Question Answer Pairs with Bilateral Contrastive Data Augmentation6
A novel and easy-to-apply data augmentation strategy, namely Bilateral Generation (BiG), with a contrastive training objective for improving the performance of ranking question answer pairs with existing labeled data.
- Word Embedding as Maximum A Posteriori Estimation6
This paper considers parametrized variants of the GloVe model and incorporates priors on these parameters, and proposes a word embedding model in which each context word is associated with a corresponding variance, intuitively encoding how informative it is.
- Gene Ontology Classification of Biomedical Literatures Using Context Association6
It is shown that incorporation of context association can enhance the performance of the annotation hierarchy classification problem and achieve an F-score of 60.24%, which outperforms the submission runs of TREC Genomics 2004 annotation hierarchy subtask.
- A New Approach for Semi-supervised Online News Classification6
This work develops a technique which makes use of unlabeled documents since those documents can be easily collected, such as online news from the Web, and exploits the inherent structure in the set of positive training documents guided by the provided concepts of the category.
- Lazy Learning by Scanning Memory Image Lattice6
SMILE (Scanning Memory Image LatticE) is a lazy learning framework based on a memory image lattice scanning technique that utilizes subsets of attribute values as much as possible to lead to a more flexible model which is less sensitive to data.
- Web-based Workflow Framework with CORBA6
This paper distill out the fundamental elements of a workflow system into a general workflow framework that adheres to the CORBA architecture, thus allowing tasks and resources of different types to interoperate in the framework.
- Intelligent content-based document delivery via automatic filtering profile generation6
An intelligent document delivery approach for filtering text information via a machine learning technique which automatically constructs a filtering profile from training examples that is able to operate without any prior information or restriction of the topic areas and yet achieve the filtering task.
- 5
- Dynamic Topic Tracker for KB-to-Text Generation5
This paper proposes a novel dynamic topic tracker that learns a global hidden representation for topics and recognizes the corresponding topic during each generation step and alleviates the off-topic problem.
- 5
- 5
- Structured positional entity language model for enterprise entity retrieval5
This work develops a structured positional entity language model suitable for ranking entities extracted from the webpage content incorporating the structured content representation, which captures hierarchical information blocks and semi-structured data records information.
- A Nonparametric N-Gram Topic Model with Interpretable Latent Topics5
This work presents a new nonparametric topic model that not only maintains the word order in the topic discovery process, but also generates topical n-gram words leading to more interpretable latent topics in the family of the non parametric topic models.
- An unsupervised technical difficulty ranking model based on conceptual terrain in the latent space5
An unsupervised conceptual terrain model using Latent Semantic Indexing (LSI) for re-ranking search results obtained from a similarity based search system that connects the sequences of terms under the latent space by the semantic distance between the terms and compute the traversal cost for a document indicating the technical difficulty.
- 5
- Effective fusion-based approaches for recommender systems5
A relational fusion approach based on conditional random fields (CRF) is proposed to improve traditional fusion methods by incorporating global relational dependency and to formulate the issue under the classical secretary problem framework and extend an online secretary algorithm to solve it.
- An unsupervised ranking method based on a technical difficulty terrain5
A query-independent ranking system is proposed for technical web pages that outperforms traditional readability measures.
- Learning ontology resolution for document representation and its applications in text mining5
An Adaptive Concept Resolution (ACR) model is proposed that can learn a concept border from an ontology taking into consideration of the characteristics of a particular document collection and provide a tailor-made semantic concept representation for a document coming from the same domain.
- Lazy Learning for Classification Based on Query Projections5
A novel lazy learning method called QPAL is proposed that attempts to discover useful patterns known as query projections, which are customized to the query instance, which can guarantee to discover high-quality QPs in the learning process.
- AUTOMATIC DOMAIN ONTOLOGY GENERATION FROM WEB SITES5
A novel framework for automatically generating the domain knowledge by analyzing different Web sites in a given domain by measuring the similarity between the concepts in different Web Sites is developed.
- 5
- Automatic event generation from multi-lingual news stories5
A novel approach for automatic generation of topically-rela ted events from multi-lingual news sources by employing transformation-based linguistic tagging approach for named entity extraction.
- Chinese document indexing based on a new partitioned signature file: Model and evaluation5
This article investigates the use of signature files in Chinese information retrieval system and proposes a new partitioning method for Chinese signature file based on the characteristic of Chinese words, called Partitioned Signature File for Chinese (PSFC), which offers faster search efficiency than the traditional single signature file approach.
- Abstraction in Bayesian belief networks and automatic discovery from past inference sessions5
An abstraction scheme is developed to simplify Bayesian belief network structures for future inference sessions and an approach for automatic discovery of good abstractions from the past inference sessions is presented.
- Unveiling the Entropy Dynamics of Chain-of-Thought Reasoning4
This paper investigates the entropy dynamics of Chain-of-Thought (CoT) and uncovers a consistent two-phase structure: an Uncertainty Region of exploration transitioning sharply to a Confidence Region of convergence, and develops a training-free framework for real-time inference control.
- Unveiling Knowledge Boundary of Large Language Models for Trustworthy Information Access4
This tutorial will explore the taxonomy of knowledge boundary in LLMs, addressing their handling of uncertainty, response calibration, and mitigation of unintended behaviors that can arise during interaction with users.
- 4
- Parameter-Efficient Tuning by Manipulating Hidden States of Pretrained Language Models For Classification Tasks4
This paper proposes a simple tuning method which only introduces three trainable vectors and achieves comparable performance with prompt tuning methods like P- Tuning and P-tuning v2, verifying that original hidden states do contain useful information for classification tasks.
- 4
- Unsupervised KB-to-Text Generation with Auxiliary Triple Extraction using Dual Learning4
This paper proposes a method which trains the generation model in a completely unsupervised way with unaligned raw text data and KB triples through a novel dual training framework which leverages the inverse relationship between the KB-to-text generation task and an auxiliary triple extraction task.
- 4
- Answering Product-related Questions with Heterogeneous Information4
A novel framework to tackle the PQA task via exploiting heterogeneous information including natural language text and attribute-value pairs from two information sources of the concerned product, namely product details and user reviews is proposed.
- Towards a language-independent solution: Knowledge base completion by searching the Web and deriving language pattern4
A language-independent framework to tackle the slot-filling task by searching the Web with high-precision queries, and deriving lightweight extraction patterns based on string matching, which make no use of complex NLP resources, which may be unavailable in some languages.
- Knowledge Base Refinement Using Limited Amount of Efforts from Experts4
The authors propose a new knowledge base refinement framework based on pattern mining and active learning using an existing available knowledge base constructed from a different domain (source domain) solving the same task as well as some data collected from the target domain.
- 4
- Bidirectional Integration of Pipeline Models4
This paper introduces joint factors connecting variables of relevant subtasks to capture dependencies and interactions between them and proposes a strong bidirectional MCMC sampling inference algorithm which allows information to flow in both directions to find the approximate MAP solution for all subtasks.
- An unsupervised approach for product record normalization across different web sites4
An unsupervised probabilistic learning framework for normalizing product records across different retailer Web sites is presented and a graphical model for the generation of text fragments in Web pages is developed to accomplish the two tasks.
- Utilizing hierarchical feature domain values for prediction4
A Bayesian learning framework which can exploit hierarchical structures of discrete feature domain values to improve the prediction performance on sparse training data is proposed and a tradeoff between precision and robustness is achieved to improved the parameter estimation for prediction.
- 4
- 4
- Meta-learning Models for Automatic Textual Document Categorization4
The second meta-model approach is proposed, called Meta-learning Using Document Feature characteristics (MUDOF), which employs a meta-learning phase using document feature characteristics.
- Using discretization and Bayesian inference network learning for automatic filtering profile generation4
A new approach for text document filtering based on automatic construction of filtering profiles using Bayesian inference network learning with advanced discretization achieves better performance over the simple naive Bayesian approach.
- EPA: Easy Prompt Augmentation on Large Language Models via Multiple Sources and Multiple Targets3
A novel method called EPA that effectively minimizes user efforts in writing demonstrations while improving the model performance at the same time, and automatically augmenting the demonstrations with multiple sources/targets, where each of them paraphrasing each other.
- Multimodal Fashion Knowledge Extraction as Captioning3
This work proposes to cast the task of social media fashion knowledge extraction as a captioning problem to capture the interplay of the multimodal post information and transforms the fashion knowledge tuples into a natural language caption with a sentence transformation method.
- Revamping Multilingual Agreement Bidirectionally via Switched Back-translation for Multilingual Neural Machine Translation3
A novel framework called M ultilingual Trans itivity (MTrans) is proposed to provide an implicit optimal route via source-auxiliary-target within the multilingual model and a novel framework called Bi directional M ultilingual Agree ment (Bi-MAgree) that encourages the bidirectional agreement be-tween different languages.
- Domain-Aware Recurrent Neural Network for Cross-Domain Sentiment Classification3
This work proposes a new domain-aware recurrent neural network leveraging the customized domain adversarial technique to minimize the domain discrepancy to solve the problem of scarce or even no labeled data in the target domain.
- 3
- 3
- Babbelbord: A Personalized Conversational Game for People with Dementia3
How the Babbelbord was designed to offer an engaging and gamified user experience, tailored for older adults affected by dementia is described and demonstrated.
- CUIS at the NTCIR-12 MobileClick2 Task.3
This work conducts intent discovery based on latent topic modeling and exploits the discovered intents and considers the importance of an iUnit in each Web content document, and develops the iUnit summarization model using the outcome from the iUnit ranking subtask.
- Nonparametric Topic Modeling Using Chinese Restaurant Franchise with Buddy Customers3
A new metaphor in Bayesian nonparametrics known as the “Chinese Restaurant Franchise with Buddy Customers” is proposed, which shows an improvement over existing comparative models.
- Topic modeling for conference analytics3
This work presents the attempt to understand the research topics that characterize the papers submitted to a conference, by using topic modeling and data visualization techniques, and compares the automatically inferred topics against the expert-defined topics.
- A Unified Posterior Regularized Topic Model with Maximum Margin for Learning-to-Rank3
This work proposes a learning-to-rank framework which integrates the supervised learning of a maximum margin classifier with the discovery of a suitable probabilistic topic model, and uses a unified constrained optimization framework, which can dynamically compute the latent topic similarity score between the query and the document.
- Investigation of Web Query Refinement via Topic Analysis and Learning with Personalization3
This proposed framework exploits a latent topic space, which is automatically derived from a query log, to leverage the semantic dependency of terms in a query to achieve personalization of query renement.
- Exploring Query Matrix for Support Pattern Based Classification Learning3
It is shown that SUPE conducts the search from specific to general in a significantly reduced hypothesis space and the query matrix also facilitates the complicated operations in SUPE.
- Text Clustering with Limited User Feedback Under Local Metric Learning3
The results demonstrate that user feedback information coupled with local metric learning can dramatically improve the clustering performance.
- 3
- CatRelate: A New Hierarchical Document Category Integration Algorithm by Learning Category Relationships3
This work addresses the problem of integrating documents from a source catalog into a master catalog by correctly identifying relationships among categories, such as Match, Disjoint, SubCon concept, SuperConcept, and Overlap, which come from the relations of sets in Set theory.
- HIGH‐DIMENSIONAL LEARNING FRAMEWORK FOR ADAPTIVE DOCUMENT FILTERING*3
This work proposes a new high‐dimensional on‐line learning framework, known as the REPGER (relevant feature pool with good training example retrieval rule) algorithm, which maintains a pool of selective features with potentially high predictive power to predict document relevance.
- 3
- 3
- Learning via Prototype Generation and Filtering3
A new framework for integration of instance-filtering and instance-averaging techniques that shows good performance in data reduction while maintaining or even improving classification accuracy in 19 real data sets is investigated.
- An intelligent adaptive filtering agent based on an on-line learning model (poster abstract)3
This work investigates the continuous performance of the whole filtering process starting from the beginning and proposes a new on-line lesrning algorithm, known as the REPGER (RElevant term Pool with Good training Example retrieval Rule) algorithm, for this task.
- 2
- Rephrasing Invokes Better Generations for Large Language Models2
ReLLM (Rephrasing for LLMs), a method that automatically paraphrases input content for better output generations, is proposed, which is user-friendly and requires no additional LLM training.
- 2
- 2
- Functional Distributional Semantics at Scale2
This paper proposes a more expressive lexical model that works over a continuous semantic space, which improves the flexibility and computational efficiency of the model, as well as its compatibility with present-day machine-learning frameworks.
- 2
- Towards Domain-Generalizable Paraphrase Identification by Avoiding the Shortcut Learning2
This paper proposes a PI framework based on Optimal Transport (OT) and proposes a method that forces the network to learn the necessary features for all the words in the input, which alleviates the shortcut learning problem.
- 2
- Point-of-Interest Oriented Question Answering with Joint Inference of Semantic Matching and Distance Correlation2
A novel deep learning framework integrated with joint inference to capture both tag semantic and geographic correlation between question and POIs and a special cross attention question embedding neural network structure to obtain question- to-POI and POI-to-question information is presented.
- Incorporating Pseudo-Parallel Data for Quantifiable Sequence Editing.2
This work proposes a framework which performs QuaSE by incorporating pseudo-parallel data, which enables a better disentanglement of the latent factors that are relevant to the outcome and thus provides a solid basis to generate output satisfying the desired outcome.
- Entity Retrieval via Type Taxonomy Aware Smoothing2
A type taxonomy aware smoothing method that exploits the hierarchical type information of a knowledge base and integrates into an existing language modelling framework that improves the performance of existing models is proposed.
- A learning framework for information block search based on probabilistic graphical models and Fisher Kernel2
This work has developed a novel learning framework for retrieving precise information blocks from Web pages given a query, which may contain some search terms and prior information such as the layout format of the data.
- 2
- Website Community Mining from Query Logs with Two-Phase Clustering2
This paper proposes a method to automatically mine website communities by exploiting the query log data in Web search, and clusters the queries of the same host to obtain different content aspects of the host.
- Accelerated Training of Maximum Margin Markov Models for Sequence Labeling: A Case Study of NP Chunking2
This work presents the first known empirical results on sequence labeling based on maximum margin Markov networks (M3N), which incorporate both kernel methods to efficiently deal with high-dimensional feature spaces, and probabilistic graphical models to capture correlations in structured data.
- Building Knowledge Base for Reading from Encyclopedia.2
This paper presents a framework using Encyclopedia texts to induce and generalize entities and relations existing in the corpus so that the extracted elements can be used for processing more general text.
- Extracting and Summarizing Hot Item Features Across Different Auction Web Sites2
This work develops a unified framework aiming at automatically extracting the product features and summarizing the hot item features across different auction Web sites using conditional random fields.
- 2
- 2
- 2
- 2
- 1
- 1
- 1
- 1
- CO3: Low-resource Contrastive Co-training for Generative Conversational Query Rewrite1
This work co-train two dual models such that each of them provides extra guidance through pseudo-labeling for enhancing the other in an iterative manner to utilize massive unlabeled data to make further improvements via a contrastive co-training paradigm.
- 1
- 1
- Semantic Composition with PSHRG for Derivation Tree Reconstruction from Graph-Based Meaning Representations1
This work reveals the ability of PSHRG in formalizing a syntax–semantics interface, modelling compositional graph-to-tree translations, and channelling explainability to surface realization.
- Using Latent Fine-Grained Sentiment for Cross-Domain Sentiment Analysis1
A novel model is proposed which is capable of capturing both sentiment terms with strong or weak polarity from various source domains which are useful for knowledge transfer to unlabeled target domain and has a good performance even under the little labeled data constraint.
- 1
- 1
- Intra-/Inter-Interaction Network with Latent Interaction Modeling for Multi-turn Response Selection1
This work proposes Intra-/Inter-Interaction Network (I^3) with latent interaction modeling to comprehensively model multi-level interactions between the utterance context and the response.
- CUIS Team at TREC 2018 CAR Track.1
This work proposes a Markov random field based framework concerning unigrams, bigrams and concepts from differnt query sections and employs a language modelling framework facilitating the Wikipedia article information and query entity mentions.
- Digesting News Reader Comments via Fine-Grained Associations with Event Facets and News Contents1
An unsupervised model called DRC is proposed, based on collective matrix factorization and develop a multiplicative-update method to infer the parameters and Experimental results show that the proposed DRC model can provide an effective way to digest news reader comments.
- Exploring Urban Lifestyles Using a Nonparametric Temporal Graphical Model1
This work proposes a new unsupervised nonparametric temporal topic model to discover lifestyle patterns from location-based social networks by relating the textual content, time stamps, and venue categories associated to user check-in to detect the predominant lifestyle patterns in a given geographic region.
- Learning a Semantic Space of Web Search via Session Data1
session2vec and session2vec+ models are proposed to learn vectors in the space with search session data, where a search session is regarded as an instantiation of an information need and keeps the interaction information of queries and URLs.
- Hidden Aspect Rating Discovery from Text Reviews of E-Commerce Web Sites1
Experimental results on the two real-world product review corpora demonstrate that the proposed generative model outperforms existing state-of-the-art models in the aspect rating prediction task.
- Logic Relation Refinement Using Unlabeled Data1
This work has developed a framework for the adaptation of relational logic models, in particular, Markov Logic Network (MLN), from a source domain to a target domain solving the same task using only unlabeled data in the target domain.
- A Semi-Supervised Framework for Feature Mapping and Multiclass Classification1
It is shown that this problem is equivalent to a sequential convex optimization problem by applying constraint concave-convex procedure (CCCP) and efficient algorithm with theoretical guarantee is proposed and computational issue is investigated.
- Cross Language Information Extraction Knowledge Adaptation1
The idea of the framework is to utilize the previously learned information extraction knowledge and the previously extracted or collected items in the source Web site to utilizing the new wrapper for the new unseen site using the language independent wrapper induction component.
- Coreference resolution using expressive logic models1
A framework of coreference resolution based on first-order logic and probabilistic graphical model, the Markov Logic Network is proposed, which enables the use of background knowledge and captures more complex coreference linkage properties through rich expression of conditions.
- 1
- An Online Cascaded Approach to Biomedical Named Entity Recognition1
An online cascaded approach to biomedical named entity recognition that uses an online training method to substantially reduce the training time required and a cascaded framework to relax the memory requirement is presented.
- Introduction to the special topic section on mining Web resources for enhancing information retrieval1
This special topic section focuses on advanced techniques for mining Web resources to support and/or enhance information retrieval or seeking processes by exploiting different kinds of Web resources.
- Exploiting Hierarchical Domain Values in Classification Learning1
A framework which can exploit hierarchical structures of feature domain values to improve classification performance is proposed and a tradeoff between precision and robustness is achieved to improve the parameter estimation in classification learning.
- Digesting Bilingual Web News via Mining Financial Activities and Events1
A financial knowledge management system, known as FAM (Financial Activity Mining), which is able to digest online multilingual news and conduct financial activity mining, which is particularly helpful for tracking stock performance of a targeted company or an event.
- Beyond Supervised Learning of Wrappers for Extracting Information from Unseen Web Sites1
The problem of wrapper adaptation which aims at adapting a previously learned wrapper to an unseen target Web site is investigated, and the problem of training example identification for the target site as a hybrid text classification problem is posed.
- 1
- 1
- 1
- A two-level approach to learning in nonstationary environments1
Simulation studies empirically demonstrate the clear superiority of the two-level approach over the single-level learning in nonstationary environments.
- Object Recognition with Adaptive Decision Trees1
This paper describes a system which has been developed based on conjecture that humans learn to recognize objects by incrementally modifying and extending an internal representation based on the characteristics which distinguish objects from the rest of the environment.
- –
- –
- –
- –
- –
- –
- –
- –
- Informative Text Generation from Knowledge Triples–
A novel memory augmented generator that employs a memory network to memorize the useful knowledge learned during the training and utilizes such information together with the input triples to generate text in the operational or testing phase is proposed.
- Exploiting Domain-Aware Aspect Similarity for Multi-Source Cross-Domain Sentiment Classification–
This article is intended to provide a history of the project and some of the key events leading up to and during the development of the plan.
- –
- Sentence Structure andWord Relationship Modeling for Emphasis Selection–
A new framework that considers sentence structure via a sentence structure graph and word relationship via a word similarity graph is proposed that can achieve superior performance in emphasis selection.
- –
- Soft Kernel-based Ranking on a Statistical Manifold–
This work proposes a kernel-based neural ranking model based on a statistical manifold that considers the interaction as geodesic on a manifold and proposes a smoothed kernel pooling scheme at different similarity levels based on Riemann normal distribution.
- Fact Discovery from Knowledge Base via Facet Decomposition–
A novel framework is proposed that decomposes the discovery problem into several facet discovery components and contains a novel auto-encoder based facet component to estimate some facets of the fact.
- Guiding Approximate Text Classification Rules via Context Information–
The results demonstrate that the proposed framework performs better than existing models in some benchmarking datasets, indicating that learning the context of rules is constructive for improving text classification performance.
- –
- Digesting Multilingual Reader Comments via Latent Discussion Topics with Commonality and Specificity–
A new model called TDCS is proposed based on graphical models, which can cope with the language gap and detect language-common and language-specific latent discussion topics simultaneously simultaneously.
- A Joint Framework for Collaborative Filtering and Metric Learning–
A framework for jointly conducting collaborative filtering and distance metric learning based on regularized singular value decomposition (RSVD), which discovers the user matrix and item matrix in the low rank space, which achieves a promising prediction performance and outperforms the existing works.
- Abstract Venue Concept Detection from Location-Based Social Networks–
A new graphical model that can generate latent abstract concepts of venues, or Point of Interest (POI) by exploiting text data in venue profiles obtained from location-based social networks (LBSNs).
- Information Discovery from Semi-Structured Record Sets on the Web–
This thesis presents a record segmentation search tree framework in which a new search structure, named Record Segmentation Tree (RST), is designed and several efficient search pruning strategies on the RST structure are proposed to identify the records in a given Web page.
- –
- Mining a shared concept space for domain adaptation in text mining–
A novel model to learn a low-rank shared concept space with respect to two criteria simultaneously: the empirical loss in the source domain, and the embedded distribution gap between the sourcedomain and the target domain is developed.
- –
- Entity discovery by exploiting contextual structures–
This thesis presents a cascaded framework for extracting named entities from text documents that automatically derive features on a set of documents from different feature templates, and develops three models: maximum-entropy margin-based (MEMB), isomeric conditional random field (ICRF) model, and online cascaded reranking (OCR) model.
- –
- Logic knowledge base refinement using unlabeled or limited labeled data–
A new framework of refining a kind of logic knowledge base known as Markov Logic Networks (MLN) is investigated, showing consistent improvements in both situations of using unlabeled target domain data, and with a limited amount of labeled targetdomain data.
- –
- Discovery of logic relations for text mining adaptation using unlabeled data–
This work develops a framework to discover new logic formulae for the adaptation of first-order logic knowledge from a source domain to a target domain for solving the same task using unlabeled data only.
- –
- An Efficient Method for Generating, Storing and Matching Features for Text Mining–
A new algorithmic framework called F-tree is proposed for automatically generating and storing conjunctions of features in text mining tasks and a systematic method for removing redundant features is proposed to further reduce memory usage and improve performance.
- Semi-supervised document clustering with active learning–
A novel framework for automatically discovering pairwise constraints revealing the user grouping preference is developed for semi-supervised document clustering taking into consideration of constraints given by users.
- Exploring attributes and instances for customized learning based on support patterns–
A novel customized learning framework called Customized Support Pattern Learner (CSPL), which exploits a tradeoff between instance- based learning and attribute-based learning, and demonstrates that learning from specific to general can discover more useful patterns for learning.
- Training Conditional Random Fields with Unlabeled Data and Limited Number of Labeled Examples–
This work attempts to extend the model so that human effort in preparing labeled training examples can be reduced by considering unlabeled data, and aims at maximizing the likelihood of the observation of the sequences from both of the labeled and unlabeling data.
- Integrating XML and CORBA to support collaborative writing using off-the-shelf editing software–
A data definition scheme is proposed whereby documents are transformed into the XML format for internal storage and a hierarchical structure is superimposed on each document, allowing it to be broken down into a number of independent parts.
- –
- –
- Discovering Patterns from Ontology-Derived Texts–
A framework for constructing semantic features for textual documents from tackling the problem of abstracting information in document representation is proposed and semantic patterns discovered from ontology-derived texts are discovered.
- Interpretable Query Projection Learning–
This work proposes a novel lazy learning method, called QPL, which attempts to discover useful patterns from training data, which is customized to the query instance and easily interpretable.
- –
- –
- –
- Exploiting Heterogeneous Features for Classification Learning–
A framework for handling heterogeneous features containing hierarchical values and texts under Bayesian learning is proposed that makes use of a statistical technique called shrinkage and an approach for utilizing text data to improve classification performance.
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.