Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks121 from public data
- Adapting ranking SVM to document retrieval599
Experimental results show that the modifications made in conventional Ranking SVM can outperform the conventional ranking SVM and other existing methods for document retrieval on two datasets and employ two methods to conduct optimization on the loss function: gradient descent and quadratic programming.
- Low-Quality Product Review Detection in Opinion Summarization424
Experimental results show that the proposed approach effectively discriminates lowquality reviews from high-quality ones and enhances the task of opinion summarization by detecting and filtering low quality reviews.
- Email data cleaning171
A cascaded approach is proposed, which cleans up an email in four passes including non-text filtering, paragraph normalization, sentence normalization, and word normalization, which becomes independent from any specific text mining processing.
- Searching Questions by Identifying Question Topic and Question Focus170
Experimental results indicate that the approach of identifying question topic and question focus for search significantly outperforms the baseline methods such as Vector Space Model (VSM) and Language Model for Information Retrieval (LMIR).
- Understanding and summarizing answers in community-based question answering services125
An in-depth analysis of questions and answers on cQA services finds that the assumption that questions always have unique best answers cannot be true, and shows that question-type oriented summarization techniques can improve cZA answer quality significantly.
- Read, Listen, and See: Leveraging Multimodal Information Helps Chinese Spell Checking118
The ReaLiSe model tackles the CSC task by capturing the semantic, phonetic and graphic information of the input characters, and selectively mixing the information in these modalities to predict the correct output.
- Research on Expert Search at Enterprise Track of TREC 2006118
In this research, a new evidence-oriented framework to expert search is proposed, whereby the task of Expert Search can be accomplished in three steps, namely, 1) evidence extraction, 2) evidence quality evaluation, and 3) evidence merging.
- 117
- Base Noun Phrase translation using web data and the EM algorithm116
Experimental results indicate that the coverage and accuracy of the method are significantly better than those of the baseline methods relying on existing technologies.
- 95
- Making Large Language Models Better Reasoners with Alignment94
The constraint alignment loss has two objectives: a) Alignment, which guarantees that positive scores surpass negative scores to encourage answers with high-quality COTs, and b) Constraint, which keeps the negative scores confined to a reasonable range to prevent the model degradation.
- 91
- 88
- An Enhanced Span-based Decomposition Method for Few-Shot Sequence Labeling73
SD decomposes the span matching problem into a series of span-level procedures, mainly including enhanced span representation, class prototype aggregation and span conflicts resolution, which achieves the new state-of-the-art results on two popular FSSL benchmarks, FewNERD and SNIPS.
- Asking Effective and Diverse Questions: A Machine Reading Comprehension based Framework for Joint Entity-Relation Extraction70
This paper improves the existing MRC-based entity-relation extraction model through diverse question answering and significantly outperforms baseline models, which improves the relation F1 to 62.1% on ACE05 and 71.9% on CoNLL04.
- Knowledge-Sensed Cognitive Diagnosis for Intelligent Education Platforms68
A Knowledge-Sensed Cognitive Diagnosis (KSCD) framework is proposed, aiming at learning intrinsic relations among knowledge concepts from student response logs and incorporating them for inferring students' mastery over all knowledge concepts in an end-to-end manner.
- Recommending questions using the mdl-based tree cut model68
Experimental results indicate that the use of the MDL-based tree cut model can significantly outperform the baseline methods of word- based VSM or phrase-based VSM.
- Title extraction from bodies of HTML documents and its application to web page retrieval67
Experimental results indicate that the use of both extracted titles and title fields is almost always better than theuse of title fields alone; theUse of extracted titles is particularly helpful in the task of named page finding.
- Seeking Patterns, Not just Memorizing Procedures: Contrastive Learning for Solving Math Word Problems62
This paper investigates how a neural network understands patterns only from semantics, and proposes a contrastive learning approach, where the neural network perceives the divergence of patterns.
- Collective Tweet Wikification based on Semi-supervised Graph Regularization59
A novel semi-supervised graph regularization model is proposed to incorporate both local and global evidence from multiple tweets through three fine-grained relations in order to identify semanticallyrelated mentions for collective inference.
- Ranking definitions with supervised learning methods59
Experimental results indicate that the use of SVM and Ranking SVM can significantly outperform the baseline methods of using heuristic rules or employing the conventional information retrieval method of Okapi, indicating that generic models for definition ranking can be constructed.
- Enhancing Label Correlation Feedback in Multi-Label Text Classification via Multi-Task Learning54
This work introduces a novel approach with multi-task learning to enhance label correlation feedback and proposes two auxiliary label co-occurrence prediction tasks to enhancelabel correlation learning.
- Improving BERT with Syntax-aware Local Attention52
A syntax-aware local attention, where the attention scopes are restrained based on the distances in the syntactic structure is proposed, which can be integrated with pretrained language models, such as BERT, to render the model to focus on syntactically relevant words.
- Dialogue Response Selection with Hierarchical Curriculum Learning51
A hierarchical curriculum learning framework that trains the matching model in an “easy-to-difficult” scheme for dialogue response selection significantly improves the model performance across various evaluation metrics.
- Web page title extraction and its application50
Experimental results indicate that the use of both extracted titles and title fields is almost always better than theuse of title fields alone; the useof extracted titles is particularly helpful in the task of named page finding.
- Difference-aware Knowledge Selection for Knowledge-grounded Conversation Generation48
Automatic, human observational, and interactive evaluation shows that the proposed difference-aware knowledge selection method is able to select knowledge more accurately and generate more informative responses, significantly outperforming the state-of-the-art baselines.
- 47
- Automatic extraction of web data records containing user-generated content45
This work finds that certain domain constraints can be used to design better similarity measures capable of circumventing the influence of UGC, and uses anchor points provided by the domain constraints to improve the extraction process, which ends in an algorithm called MiBAT (Mining data records Based on Anchor Trees).
- A Statistical Framework for Product Description Generation39
A statistical framework that generates accurate and fluent product description from product attributes by using learned knowledge to decide what to say and how to say for product description generation is presented.
- 38
- Learning to recommend questions based on user ratings38
A new algorithm called 'majority-based perceptron algorithm' is designed which can avoid the influence of noisy instances by emphasizing its learning over data instances from the majority users.
- 37
- 36
- Cost-Sensitive Learning of SVM for Ranking35
Simulation results show that the proposed Ranking SVM model works better than ranking SVM in practical settings of ranking, and experimental results indicate that the method can outperform existing methods including Ranking S VM on real information retrieval tasks such as document search and definition search.
- Correlation-Guided Representation for Multi-Label Text Classification33
The task is viewed as a correlation-guided text representation problem: an attention-based two-step framework is proposed to integrate text information and label semantics by jointly learning words and labels in the same space to capture high-order label-label correlations as well as context- label correlations.
- RadarMath: An Intelligent Tutoring System for Math Education33
A novel intelligent tutoring system, called RadarMath, is proposed and implemented to support intelligent and personalized learning for math education and demonstrates how the relevant AI techniques could be applied in today's intelligent Tutoring systems.
- Mention and Entity Description Co-Attention for Entity Disambiguation33
A type-aware co-attention model for entity disambiguation is proposed, which tries to identify the most discriminative words from mention contexts and most relevant sentences from corresponding entity descriptions simultaneously.
- Using social annotations to improve language model for information retrieval33
Evaluations using del.icio.us data show that LAM outperforms the traditional LMIR approaches significantly, and can boost LMIR further when similarity among annotations and similarity among documents are taken into consideration simultaneously.
- Automatic Context Pattern Generation for Entity Set Expansion31
The GAPA is proposed, a novel ESE framework that leverages the aforementioned GenerAted PAtterns to expand target entities and devise a context pattern generation module that utilizes autoregressive language models (e.g., GPT-2) to automatically generate high-quality context patterns for entities.
- A Supervised Learning Approach to Entity Search29
Using a linear model to combine the uses of different features and employing a supervised learning approach in training of the model significantly outperforms the baseline method based solely on co-occurrences in entity search.
- 28
- Learning to Suggest Questions in Online Forums28
This paper presents a new framework to suggest questions, and proposes the Topicenhanced Translation-based Language Model (TopicTRLM) which fuses both the lexical and latent semantic knowledge.
- 27
- A new approach to intranet search based on information extraction27
This paper proposes a new approach to intranet search in which the authors search for information in each of the special types, in addition to the traditional relevance search, and develops an intranets search system called 'Information Desk'.
- Contextual Similarity is More Valuable Than Character Similarity: An Empirical Study for Chinese Spell Checking24
A simple yet effective Curriculum Learning (CL) framework for the Chinese Spell Checking task, which empirically suggests that contextual similarity is more valuable than character similarity for the CSC task.
- Learning Robust Representations for Continual Relation Extraction via Adversarial Class Augmentation24
- A Prerequisite Attention Model for Knowledge Proficiency Diagnosis of Students22
A Prerequisite Attention model for Knowledge Proficiency diagnosis of students (PAKP) is proposed to learn the attentive weights of precursor concepts on successor concepts and model it for inferring the knowledge proficiency.
- A Unified Multi-Task Learning Framework for Joint Extraction of Entities and Relations22
A unified multi-task learning framework to divide the joint extraction of entities and relations into three interacted sub-tasks and introduces the type-attentional method for subject extraction to provide prior type information explicitly and a question generation based QA method for object extraction to obtain diverse queries automatically.
- A structural support vector method for extracting contexts and answers of questions from online forums22
A novel and unified model is proposed by customizing the structural Support Vector Machine method that gives a comprehensive graphical representation of thread discussion and designs special inference algorithms instead of general-purpose ones.
- 18
- I2R-NUS-MSRA at TAC 2011: Entity Linking.18
The joint participation of I2R-NUS team and MSRA team in entity linking task for Knowledge Base Population at Text Analysis Conference 2011 is reported.
- Research on Enterprise Track of TREC 2007 at SJTU APEX Lab18
Among all the techniques embedded in this system, the highlight is the static ranking approaches and some data preprocessing methods and similarity function will also be introduced.
- Hierarchical Curriculum Learning for AMR Parsing17
A Hierarchical Curriculum Learning (HCL) framework with Structure-level (SC) and Instance-level Curricula (IC) that reduces the difficulty of learning complex structures, thus the flat model can better adapt to the AMR hierarchy.
- 16
- 14
- Pre-training and Fine-tuning Neural Topic Model: A Simple yet Effective Approach to Incorporating External Knowledge14
Results show that the proposed approach significantly outperforms both current state-of-the-art neural topic models and some topic modeling approaches enhanced with PWEs or PLMs and greatly reduces the need for the huge size of training data.
- 14
- Pretraining without wordpieces: learning over a vocabulary of millions of words12
This work explores to develop a BERT-style pretrained model over a vocabulary of words instead of wordpieces, called WordBERT, which makes significant improvements on cloze test and machine reading comprehension and has comparable time cost to BERT in natural language understanding tasks.
- 12
- 12
- Less is More: Rethinking State-of-the-art Continual Relation Extraction Models with a Frustratingly Easy but Effective Approach10
This paper proposes a Frustratingly easy but Effective Approach (FEA) method with two learning stages for CRE, and finds that two strong CRE baselines can be subsumed into the proposed training pipeline.
- Exploiting Unlabeled Data via Partial Label Assignment for Multi-Class Semi-Supervised Learning10
Comparative studies against state-of-the-art approaches clearly show the effectiveness of the proposed unlabeled data exploitation strategy for multi-class semi-supervised learning.
- Searching Documents Based on Relevance and Type10
Experimental results show the proposed approaches can significantly outperform the baseline methods on the problem of 'typed search'.
- Instance Segmentation for Chinese Character Stroke Extraction, Datasets and Benchmarks9
This work presents the first public available Chinese Character Stroke Extraction ( CCSE) benchmark, with two new large-scale datasets: Kaiti CCSE (CCSE-Kai) and Handwritten CCSE-HW, and turns the stroke extraction problem into a stroke instance segmentation problem.
- A Non-Hierarchical Attention Network with Modality Dropout for Textual Response Generation in Multimodal Dialogue Systems9
A non-hierarchical attention network with modality dropout is proposed, which abandons the HRED framework and utilizes attention modules to encode each utterance and model the context representation.
- 9
- Diversity and Consistency: Exploring Visual Question-Answer Pair Generation8
This paper introduces a novel task that targets question-answer pair generation from visual images and proposes three models: the pipeline model, the joint model, and the sequential model, which can generate diverse or consistent pairs.
- Enhancing Dialogue-based Relation Extraction by Speaker and Trigger Words Prediction8
This paper proposes two beneficial tasks, speaker prediction and trigger words prediction, to enhance the extraction of dialogue-based relations, and shows noticeable improvements compared to the baseline models.
- Uncertainty reduction in collaborative bootstrapping8
A new measure for representing the degree of uncertainty correlation of the two classifiers in collaborative bootstrapping is proposed and uses the measure in analysis of collaborativebootstrapping.
- G4: Grounding-guided Goal-oriented Dialogues Generation with Multiple Documents7
A three-stage solution, Grounding-guided goal-oriented dialogues generation (G4), which predicts groundings from retrieved passages to guide the generation of the final response of the MultiDoc2Dial task.
- 6
- Entity Relative Position Representation Based Multi-head Selection for Joint Entity and Relation Extraction6
An entity-specific Relative Position Representation (eRPR) is proposed to allow the model to fully leverage the distance information between entities and context tokens and improve the semantic representation by adopting a pre-trained language model BERT as the feature encoder.
- Microsoft Research Asia with Redmond at the NTCIR-8 Community QA Pilot Task6
The approaches that were used for the NTCIR-8 Community QA Pilot task were described and the official evaluation results showed that the proposed features and learning approaches are effective in terms of finding the best quality answers.
- A Probabilistic Model for Fine-Grained Expert Search6
This paper proposes conducting expert search with a fine-grained approach, and utilizes more specific evidences existing in the documents, an evidence-oriented probabilistic model for expert search and a method for the implementation.
- A Two-Stage Model for Expert Search6
New experimental results on two-stage model using the data in both TREC 2005 and TREC 2006 expert search tasks are reported and it is shown that the approach is applicable beyond TREC W3C corpus, by introducing experimentalresults on expert finding at Microsoft Research.
- Unifying Token- and Span-level Supervisions for Few-shot Sequence Labeling5
A consistent greedy inference algorithm that first adjusts the predicted probability and then greedily selects non-overlapping spans with maximum probability is proposed and a Consistent Dual Adaptive Prototypical (CDAP) network for few-shot sequence labeling is proposed.
- Re-ranking question search results by clustering questions5
Experimental results show that the proposed approach to question clustering improves the performance of question search significantly better than the approach not utilizing the topic–focus structure.
- 4
- 3
- 3
- 3
- A Divide-And-Conquer Approach for Multi-label Multi-hop Relation Detection in Knowledge Base Question Answering3
A novel divide-and-conquer approach for multi-label multi-hop relation detection (DCMLMH) is proposed by decomposing it into head relation detection and conditional relation path generation and a novel path sampling mechanism is proposed to generate diverse relation paths for the inference stage.
- Overview of the NLPCC 2018 Shared Task: Spoken Language Understanding in Task-Oriented Dialog Systems3
The overview for the shared task at the 7th CCF Conference on Natural Language Processing & Chinese Computing (NLPCC 2018): Spoken Language Understanding (SLU) in Task-oriented Dialog Systems provides a dictionary of values for each enumerable type of slot.
- Using Social Annotations to Smooth the Language Model for IR3
An exploration of using social annotations provided by the Web 2.0 sites in helping web search by using the social annotations as an additional resource to strengthen existing smoothing methods for the language model for IR.
- Learning and Exploiting Non-Consecutive String Patterns for Information Extraction3
This algorithm is novel in that it can learn non-consecutive patterns with constraints, which are necessary for information extraction, and employs an extended version of the so-called apriori algorithm at the pattern generation step.
- 2
- Read Then Respond: Multi-granularity Grounding Prediction for Knowledge-Grounded Dialogue Generation2
- AiM: Taking Answers in Mind to Correct Chinese Cloze Tests in Educational Applications2
This paper focuses on the Chinese cloze tests correction and proposes a multimodal approach, named AiM, which outperforms OCR-based methods by a large margin and takes samples of OCR datasets as the positive samples for this task, and develops a negative sample augmentation method to scale up the training data.
- SmartSales: Sales Script Extraction and Analysis from Sales Chatlog2
The proposed SmartSales system is empowered by the state-of-the-art conversational intelligence techniques and has been running on the Tencent Cloud to serve the sales teams from several different areas.
- 2
- A Lazy Learning Model for Entity Linking using Query-Specific Information2
A lazy learning model is proposed, which allows us to improve the learning process with the distribution information to the queried name and automatically label some relevant instances for the queried name leveraging its unambiguous synonyms.
- MSRA at TAC 2011: Entity Linking2
This work aims at answering the following question: Can an retrieval-based method do a good job for candidate generation?
- 2
- A Supervised Learning Approach to Search of Definitions2
Experimental results indicate that the use of SVM and Ranking SVM can significantly outperform the baseline methods such as heuristic rules, the conventional information retrieval—Okapi, or SVM regression, indicating that generic models for definition ranking can be constructed.
- 1
- When Rules Learn: A Self-Evolving Agent for Legal Case Retrieval1
This work proposes a self-evolving framework for rule-driven query rewriting that enhances BM25 without any parameter training, and reveals that LLM's capabilities to leverage previous experimental results and its intrinsic knowledge of rule elimination play critical roles in refining the rule set via self-evolution.
- 1
- LARQ: Learning to Ask and Rewrite Questions for Community Question Answering1
This paper proposes LARQ, Learning to Ask and Rewrite Questions, which is a novel sentence-level data augmentation method that significantly outperforms single BERT and XGBoost models, as well as a widely-used QG model (NQG).
- egoStellar: Visual Analysis of Anomalous Communication Behaviors from Egocentric Perspective1
EgoStellar is a visual analytics system to interactively explore the communication behaviors of mobile users from an ego network perspective, and its front-end interactive visualizations are intuitive and useful for analysts to discover insights in data.
- Learning a Replacement Model for Query Segmentation with Consistency in Search Logs1
In LRMC, a new paradigm of learning a replacement model with consistency (LRMC), an implementation of any existing approach is proposed to enable unsupervised training with guidance from search log data.
- An error driven approach to query segmentation1
An error-driven approach to query segmentation (EDQS) is proposed with the help of search logs, which enables unsupervised training with guidance from the system-specific errors.
- 1
- 1
- DisCal: Distribution-Aware Calibration for Mathematical Reasoning Under Character-Level Noisy Inputs–
Experiments demonstrate that DisCal consistently outperforms existing calibration methods under noisy inputs, reducing Expected Calibration Error (ECE) and improving Area Under the Receiver Operating Characteristic Curve (AU-ROC) by up to 31.44%.
- Grounding Multi-Hop Reasoning in Structural Causal Models via Group Relative Policy Optimization–
This work introduces an SCM-inspired framework that grounds reasoning in explicit directed dependency graphs, treating verification as a constructive structural reasoning process rather than full causal inference with interventions or counterfactual semantics.
- SHARP: Self-adaptive Harmful Category-aware Prompt Generation for Black-box Jailbreaking–
This work proposes the first category-aware jailbreak framework, SHARP, which incorporates the semantic category of harmful questions into prompt generation and achieves better cross-category robustness compared to the state-of-the-art (SOTA) baselines, providing an efficient and scalable tool for evaluating LLM safety.
- –
- –
- –
- SkillNet-X: A Multilingual Multitask Model with Sparsely Activated Skills–
A general multilingual multitask model, named SkillNet-X, which enables a single model to tackle many different tasks from different languages, and shows that SkillNet-X performs better than task-specific and two multitask learning baselines.
- IEKM: A Model Incorporating External Keyword Matrices–
An incorporation external keywords matrices model (IEKM) is proposed that uses external tools or dictionaries to construct external matrices and fuses them to the self-attention layers of the Transformer structure through gating units, thus enabling flexible corrections to the model results.
- TQ-Net: Mixed Contrastive Representation Learning For Heterogeneous Test Questions–
This paper improves previous text-only representation with a two-stage unsupervised instance level contrastive based pre-training method (MCL: Mixture Unsupervised Contrastive Learning), and proposes TQ-Net, a model to fuse the content of images to the representation of heterogeneous data.
- Bi-Drop: Enhancing Fine-tuning Generalization via Synchronous sub-net Estimation and Optimization–
Bi-Drop, a fine-tuning strategy that selectively updates model parameters using gradients from various sub-nets dynamically generated by dropout, demonstrates excellent generalization ability and robustness for domain transfer, data imbalance, and low-resource scenarios.
- Finding Similar Exercises in Retrieval Manner–
This paper provides solutions to solve the problem of low-quality annotated data and comprehensive representation of the semantic information of exercises was obtained through representation learning.
- –
- –
- MMKE: A Multi-Model Knowledge Extraction System from Unstructured Texts–
A Multi-Model Knowledge Extraction System which consists of two unstructured text extraction models (RelationSO model and SubjectRO model) based on a multi-task learning framework that incorporates a number of user-friendly extraction functionalities, such as multi-format uploading, one-click extractions, knowledge editing and graphical displays.
- –
- –
- –
- UHSIOV:A New Approach to Servo Info Verifying for Hard Disks–
Experimental results indicate that this method guarantees the integrality of Servo info verifying and shortens the time of using Servo Track Writer and verifying procedure, and compared to the traditional method, realizes unmanned verifying of servo info.
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 2026-10-15. 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.