Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks462 from public data
- ArnetMiner2,609
The architecture and main features of the ArnetMiner system, which aims at extracting and mining academic social networks, are described and a unified modeling approach to simultaneously model topical aspects of papers, authors, and publication venues is proposed.
- ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools1,810
This report primarily focuses on the GLM-4 language series, which includes GLM-4, GLM-4-Air, and GLM-4-9B, which represent the most capable models that are trained with all the insights and lessons gained from the preceding three generations of ChatGLM.
- Parameter-efficient fine-tuning of large-scale pre-trained language models1,311
Different methods of fine-tuning with only a small number of parameters are compared on a large set of natural language processing tasks and demonstrates that large-scale models could be effectively stimulated by the optimization of a few parameters.
- KEPLER: A Unified Model for Knowledge Embedding and Pre-trained Language Representation883
A unified model for Knowledge Embedding and Pre-trained LanguagERepresentation (KEPLER), which can not only better integrate factual knowledge into PLMs but also produce effective text-enhanced KE with the strong PLMs is proposed.
- CogVLM: Visual Expert for Pretrained Language Models864
This work introduces CogVLM, a powerful open-source visual language foundation model that bridges the gap between the frozen pretrained language model and image encoder by a trainable visual expert module in the attention and FFN layers, enabling deep fusion of vision language features without sacrificing any performance on NLP tasks.
- GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models474
GLM-4.5 is presented, an open-source Mixture-of-Experts (MoE) large language model with 355B total parameters and 32B activated parameters, featuring a hybrid reasoning method that supports both thinking and direct response modes that ranks 3rd overall among all evaluated models and 2nd on agentic benchmarks.
- 474
- Knowledgeable Prompt-tuning: Incorporating Knowledge into Prompt Verbalizer for Text Classification456
This work focuses on incorporating external knowledge into the verbalizer, forming a knowledgeable prompt Tuning (KPT), to improve and stabilize prompttuning.
- OpenKE: An Open Toolkit for Knowledge Embedding376
An open toolkit for knowledge embedding, which provides a unified framework and various fundamental models to embed knowledge graphs into a continuous low-dimensional space and the embeddings of some existing large-scale knowledge graphs pre-trained by OpenKE are available.
- Multi-Channel Graph Neural Network for Entity Alignment309
A novel Multi-channel Graph Neural Network model (MuGNN) is proposed to learn alignment-oriented knowledge graph (KG) embeddings by robustly encoding two KGs via multiple channels and infer and transfer rule knowledge for completing two KG consistently.
- Understanding retweeting behaviors in social networks273
This paper proposes a factor graph model to predict users' retweeting behaviors and shows that this method can achieve a precision of 28.81% and recall of 37.33% for prediction of the retweet behaviors.
- CogVLM2: Visual Language Models for Image and Video Understanding267
The CogVLM2 family, a new generation of visual language models for image and video understanding including CogVLM2, CogVLM2-Video and GLM-4V are proposed, achieving state-of-the-art results on benchmarks like MMBench, MM-Vet, TextVQA, MVBench and VCGBench.
- Explainable and Explicit Visual Reasoning Over Scene Graphs265
The prevalent black-box neural architectures used in complex visual reasoning tasks are dismantled into the proposed eXplainable and eXplicit Neural Modules (XNMs), which advance beyond existing neural module networks towards using scene graphs --- objects as nodes and the pairwise relationships as edges --- for explainable and explicit reasoning with structured knowledge.
- Expert Finding in a Social Network255
Experimental results show that the proposed propagation-based approach can outperform the baseline approach and take into consideration both person local information and network information (e.g. relationships between persons).
- A Survey of Knowledge Enhanced Pre-Trained Language Models251
A comprehensive review of Knowledge Enhanced Pre-trained Language Models (KE-PLMs) is presented to provide a clear insight into this thriving field and introduces appropriate taxonomies respectively for Natural Language Understanding (NLU) and Natural Language Generation (NLG) to highlight these two main tasks of NLP.
- Delta Tuning: A Comprehensive Study of Parameter Efficient Methods for Pre-trained Language Models250
The theoretical principles underlying the effectiveness of delta tuning are discussed and frameworks to interpret delta tuning from the perspective of optimization and optimal control are proposed, where results on over 100 NLP tasks demonstrate a comprehensive performance comparison of different approaches.
- Benchmarking Foundation Models with Language-Model-as-an-Examiner249
A novel benchmarking framework where the LM serves as a knowledgeable examiner that formulates questions based on its knowledge and evaluates responses in a reference-free manner is proposed, allowing for effortless extensibility as various LMs can be adopted as the examiner and the questions can be constantly updated given more diverse trigger topics.
- MAVEN: A Massive General Domain Event Detection Dataset232
A MAssive eVENt detection dataset is presented, which contains 4,480 Wikipedia documents, 117,200 event mention instances, and 207 event types and shows that existing ED methods cannot achieve promising results as on the small datasets, which suggests ED in real world remains a challenging task and requires further research efforts.
- 218
- Text-enhanced representation learning for knowledge graph216
The rich textual context information in a text corpus is incorporated to expand the semantic structure of the knowledge graph and each relation is enabled to own different representations for different head and tail entities to better handle 1-to-N, N- to-1 and N-To-N relations.
- Keyword Extraction Using Support Vector Machine211
Experimental results indicate that the proposed SVM based method can significantly outperform the baseline methods for keyword extraction and the accuracy of document classification can be significantly improved by using the keyword extraction method.
- 210
- Who Influenced You? Predicting Retweet via Social Influence Locality200
A novel notion of social influence locality is proposed and two instantiation functions based on pairwise influence and structural diversity are developed, which can further leverage the network-based correlation.
- RM-Bench: Benchmarking Reward Models of Language Models with Subtlety and Style192
RM-Bench is introduced, a novel benchmark designed to evaluate reward models based on their sensitivity to subtle content differences and resistance to style biases, and it is revealed that even state-of-the-art models achieve an average performance, which falls short of random-level accuracy when faced with style bias interference.
- 191
- TransferNet: An Effective and Transparent Framework for Multi-hop Question Answering over Relation Graph189
This paper proposes TransferNet, an effective and transparent model for multi-hop QA, which supports both label and text relations in a unified framework and surpasses the state-of-the-art models by a large margin.
- Semi-supervised Entity Alignment via Joint Knowledge Embedding Model and Cross-graph Model185
This paper proposes a semi-supervised entity alignment method by joint Knowledge Embedding model and Cross-Graph model (KECG), which can make better use of seed alignments to propagate over the entire graphs with KG-based constraints.
- Prompt-learning for Fine-grained Entity Typing178
This work develops a simple and effective prompt-learning pipeline, and proposes a self-supervised strategy that carries out distribution-level optimization in prompt- learning to automatically summarize the information of entity types in the zero-shot regime.
- LongWriter: Unleashing 10,000+ Word Generation from Long Context LLMs173
AgentWrite is introduced, an agent-based pipeline that decomposes ultra-long generation tasks into subtasks, enabling off-the-shelf LLMs to generate coherent outputs exceeding 20,000 words.
- MOOCCube: A Large-scale Data Repository for NLP Applications in MOOCs168
A large-scale data repository of over 700 MOOC courses, 100k concepts, 8 million student behaviors with an external resource is presented and a prerequisite discovery task is conducted as an example application to show the potential of MOOCCube in facilitating relevant research.
- Exploring and Evaluating Attributes, Values, and Structures for Entity Alignment163
An attributed value encoder is utilized and the KG is partitioned into subgraphs to model the various types of attribute triples efficiently and achieves significant improvements in cross-lingual and monolingual datasets.
- Knowledge discovery through directed probabilistic topic models: a survey161
This paper surveys an important subclass Directed Probabilistic Topic Models (DPTMs) with soft clustering abilities and their applications for knowledge discovery in text corpora, giving basic concepts, advantages and disadvantages in a chronological order.
- 157
- 155
- KQA Pro: A Dataset with Explicit Compositional Programs for Complex Question Answering over Knowledge Base154
Experimental results show that state-of-the-art KBQA methods cannot achieve promising results on KQA Pro as on current datasets, which suggests that K QA Pro is challenging and Complex KBZA requires further research efforts.
- Prerequisite Relation Learning for Concepts in MOOCs137
This work proposes a representation learning-based method for learning latent representations of course concepts, and investigates how different features capture the prerequisite relations between concepts.
- CPM: A large-scale generative Chinese Pre-trained language model135
CPM, with 2.6 billion parameters and 100GB Chinese training data, is the largest Chinese pre-trained language model, which could facilitate several downstream Chinese NLP tasks, such as conversation, essay generation, cloze test, and language understanding.
- CLEVE: Contrastive Pre-training for Event Extraction133
CLEVE is proposed, a contrastive pre-training framework for EE to better learn event knowledge from large unsupervised data and their semantic structures (e.g. AMR) obtained with automatic parsers.
- Do Pre-trained Models Benefit Knowledge Graph Completion? A Reliable Evaluation and a Reasonable Approach130
This work highlights a more accurate evaluation setting under the open-world assumption (OWA), which manual checks the correctness of knowledge that is not in KGs and proposes a novel PLM-based KGC model named PKGC, motivated by prompt tuning.
- Mining User Generated Content123
Mining User Generated Content is the first focused effort to compile state-of-the-art research and address future directions of UGC, which explains how to collect, index, and analyze UGC to uncover social trends and user habits.
- On Transferability of Prompt Tuning for Natural Language Processing122
It is found that the overlapping rate of activated neurons strongly reflects the transferability, which suggests how the prompts stimulate PLMs is essential, and that prompt transfer is promising for improving PT.
- RAIN: Social Role-Aware Information Diffusion115
A Role-Aware INformation diffusion model (RAIN) that integrates social role recognition and diffusion modeling into a unified framework is proposed that performs much better in both micro- and macro-level prediction than several alternative methods.
- Improving Event Detection via Open-domain Trigger Knowledge113
A novel Enrichment Knowledge Distillation (EKD) model to leverage external open-domain trigger knowledge to reduce the in-built biases to frequent trigger words in annotations is proposed.
- Differentiating Concepts and Instances for Knowledge Graph Embedding112
A novel knowledge graph embedding model named TransC is proposed by differentiating concepts and instances by encodes each concept in knowledge graph as a sphere and each instance as a vector in the same semantic space.
- Bridge Text and Knowledge by Learning Multi-Prototype Entity Mention Embedding108
A novel Multi-Prototype Mention Embedding model is proposed, which learns multiple sense embeddings for each mention by jointly modeling words from textual contexts and entities derived from a knowledge base, and an efficient language model based approach to disambiguate each mention to a specific sense.
- A multi-objective evolutionary algorithm for feature selection based on mutual information with a new redundancy measure108
A new feature redundancy measurement capable of accurately estimating mutual information between features with respect to the target class (MIFS-CR) is defined, based on a relevance measure and this new redundancy measure, a multi-objective evolutionary algorithm with class-dependent redundancy for feature selection (MECY-FS).
- Keyword extraction based on tf/idf for Chinese news document108
This paper gave the specification of keywords in Chinese news documents based on analyzing linguistic characteristics of news documents and then proposed a new keyword extraction method based on tf/idf with multi-strategies that can significantly outperform the baseline method.
- Social context summarization104
A dual wing factor graph (DWFG) model is proposed, which utilizes the mutual reinforcement between Web documents and their associated social contexts to generate summaries, and an efficient algorithm is designed to learn the proposed factor graph model.
- Adapting Meta Knowledge Graph Information for Multi-Hop Reasoning over Few-Shot Relations103
A meta-based multi-hop reasoning method (Meta-KGR), which adopts meta-learning to learn effective meta parameters from high-frequency relations that could quickly adapt to few-shot relations is proposed.
- HMEAE: Hierarchical Modular Event Argument Extraction102
A Hierarchical Modular Event Argument Extraction model is proposed, to provide effective inductive bias from the concept hierarchy of event argument roles and significantly outperform the state-of-the-art baselines.
- How Do Your Friends on Social Media Disclose Your Emotions?100
This paper formally formalizes the problem and proposes a novel emotion learning method by jointly modeling images posted by social users and comments added by their friends and can distinguish those comments that are closely related to the emotion expression for an image from the other irrelevant ones.
- KoLA: Carefully Benchmarking World Knowledge of Large Language Models96
A Knowledge-oriented LLM Assessment benchmark (KoLA), in which a contrastive system is adopted for better numerical comparability across tasks and models and a unique self-contrast metric for automatically evaluating knowledge-creating ability is adopted.
- 95
- Dynamic Anticipation and Completion for Multi-Hop Reasoning over Sparse Knowledge Graph92
A novel dynamic anticipation and completion strategies are applied to a multi-hop reasoning model named DacKGR over sparse KGs that outperforms state-of-the-art baselines and alleviates the sparseness problem of KGs.
- XLore: a large-scale English-Chinese Bilingual knowledge graph91
This demo presents a large-scale bilingual knowledge graph named XLore, which has adequately solved the above problems of Wikipedia-based multilingual knowledge bases.
- Cross-lingual knowledge linking across wiki knowledge bases91
The problem of cross-lingual knowledge linking is studied and a linkage factor graph model is presented, showing that this approach can achieve a high precision of 85.8% with a recall of 88.1%.
- T1: Advancing Language Model Reasoning through Reinforcement Learning and Inference Scaling87
It is demonstrated that T1 with open LLMs as its base exhibits inference scaling behavior and achieves superior performance on challenging math reasoning benchmarks and a simple strategy to examine inference scaling, where increased inference budgets directly lead to T1's better performance without any additional verification.
- EmoBench: Evaluating the Emotional Intelligence of Large Language Models84
EmoBench is proposed, a benchmark that draws upon established psychological theories and proposes a comprehensive definition for machine EI, including Emotional Understanding and Emotional Application, including Emotional Understanding and Emotional Application.
- 80
- MOOCCubeX: A Large Knowledge-centered Repository for Adaptive Learning in MOOCs78
This work constructs MOOCCubeX, a large, knowledge-centered repository consisting of 4,216 courses, 230,263 videos, 358,265 exercises, 637,572 fine-grained concepts and over 296 million behavioral data of 3,330,294 students, for supporting the research topics on adaptive learning in MOOCs.
- 78
- Identifying Potentially Important Concepts and Relations in an Ontology77
A simple yet effective Concept-And-Relation-Ranking (CARRank) algorithm is proposed to simultaneously rank the importance of concepts and relations in an iterative manner and is proved to be convergent both in principle and by experiments.
- Are Reasoning Models More Prone to Hallucination?72
An initial understanding of the hallucination in LRMs is provided and it is found that increased hallucination of LRMs is usually associated with the misalignment between model uncertainty and factual accuracy.
- T$^3$Bench: Benchmarking Current Progress in Text-to-3D Generation72
T$^3$Bench is introduced, the first comprehensive text-to-3D benchmark containing diverse text prompts of three increasing complexity levels that are specially designed for 3D generation and proposes two automatic metrics based on multi-view images produced by the 3D contents to assess both the subjective quality and the text alignment.
- Program Transfer for Answering Complex Questions over Knowledge Bases70
This paper designs a novel two-stage parsing framework with an efficient ontology-guided pruning strategy that aims to leverage the valuable program annotations on the rich-resourced KBs as external supervision signals to aid program induction for the low-resourcing KBs that lack program annotations.
- DocEE: A Large-Scale and Fine-grained Benchmark for Document-level Event Extraction63
Experiments show that there is still a big gap between state-of-the-art models and human beings (41% Vs 85% in F1 score), indicating that DocEE is an open issue, and highlights three features: large-scale manual annotations, fine-grained argument types and application-oriented settings.
- RiMOM results for OAEI 201063
The basic alignment process and alignment strategies in RiMOM are described, and then specific techniques used for different tracks are presented, and some comments on the results are given and some future work on RiMMom is discussed.
- Temporal expert finding through generalized time topic modeling62
A novel Temporal-Expert-Topic (TET) approach based on Semantics and Temporal Information based Expert Search (STMS) for temporal expert finding, which simultaneously models conferences influence and time information is proposed.
- What Happens Next? Future Subevent Prediction Using Contextual Hierarchical LSTM61
An end-to-end model which directly takes the texts describing previous subevents as input and automatically generates a short text describing a possible future subevent, which captures the two-level sequential structure of a subevent sequence.
- RiMOM-IM: A Novel Iterative Framework for Instance Matching61
An iterative framework to fully utilize the distinctive and available matching information to improve the efficiency and control the error propagation of instance matching, RiMOM-IM (RiMMom-Instance Matching).
- Fine-Grained Entity Typing via Hierarchical Multi Graph Convolutional Networks60
Hierarchical Multi Graph Convolutional Network (HMGCN), a novel Deep Learning architecture to tackle the problem of graph-based semi-supervised classification, is proposed and significantly outperforms four state-of-the-art methods.
- Course Concept Extraction in MOOCs via Embedding-Based Graph Propagation60
This paper proposes to learn latent representations for candidate concepts via an embedding-based method and develops a graph-based propagation algorithm to rank the candidate concepts based on the learned representations.
- RiMOM results for OAEI 200860
RiMOM obtains the ontology alignment results at OAEI 2008 contest through the alignment process and more than 8 different alignment strategies integrated in RiMOM are introduced.
- Community-based topic modeling for social tagging59
A TTR-LDA-Community model which combines the Latent Dirichlet Allocation model (LDA) and the Girvan-Newman community detection algorithm with an inference mechanism is proposed and applied to data from Delicious, a popular social tagging system, over the time period of 2005-2008.
- Extraction and mining of an academic social network59
This paper developed a social network system, called ArnetMiner, based on proposed methods to the above problems, which extracted and integrated 448,470 researcher profiles and 981,599 publications after the system having been in operation for two years.
- Tree-Structured Conditional Random Fields for Semantic Annotation59
Experimental results indicate that the proposed TCRFs for hierarchical semantic annotation can significantly outperform the existing linear-chain CRF model.
- Result of ontology alignment with RiMOM at OAEI'0659
This report presents the results based on the evaluation of RiMOM, a system that intends to combine multiple strategies, aiming at finding the "optimal" alignments from the source ontology to the target one.
- 58
- When does In-context Learning Fall Short and Why? A Study on Specification-Heavy Tasks57
It is found that ICL falls short of handling specification-heavy tasks, which are tasks with complicated and extensive task specifications, requiring several hours for ordinary humans to master, such as traditional information extraction tasks.
- Learning from Miscellaneous Other-Class Words for Few-shot Named Entity Recognition54
A novel model, Mining Undefined Classes from Other-class (MUCO), that can automatically induce different unde- fined classes from the other class to improve few-shot NER and improve the dis- criminative ability of NER classifier andance the understanding of predefined classes with stand-by semantic knowledge.
- Answering Complex Logical Queries on Knowledge Graphs via Query Computation Tree Optimization53
This work proposes QTO (Query Computation Tree Optimization) that can efficiently find the exact optimal solution by a forward-backward propagation on the tree-like computation graph, i.e., query computation tree.
- Actively Learning Ontology Matching via User Interaction53
This work proposes an active learning framework for ontology matching, which tries to find the most informative candidate matches to query the user and can significantly improve the matching accuracy.
- Adaptive online event detection in news streams52
This paper presents a novel document representation method based on word embeddings, which reduces the dimension and alleviates the sparse semantics compared to TF-IDF, and thus improves the efficiency and accuracy of event detection.
- AGENTIF: Benchmarking Instruction Following of Large Language Models in Agentic Scenarios49
This paper introduces AgentIF, the first benchmark for systematically evaluating LLM instruction following ability in agentic scenarios, and uses AgentIF to systematically evaluate existing advanced LLMs.
- A Mixture Model for Expert Finding48
A mixture model based on Probabilistic Latent Semantic Analysis (PLSA) is proposed to estimate a hidden semantic theme layer between the terms and the support documents of candidate experts to capture the semantic relevance between the query and the experts.
- A constraint-based probabilistic framework for name disambiguation48
A constraint-based probabilistic model for semi-supervised name disambiguation is given and Experimental results on the researcher social network show that the proposed framework significantly outperforms the baseline method using unsupervised hierarchical clustering algorithm.
- MM-MATH: Advancing Multimodal Math Evaluation with Process Evaluation and Fine-grained Classification46
This paper introduces a novel benchmark, MM-MATH, consisting of 5,929 open-ended middle school math problems with visual contexts, with fine-grained classification across difficulty, grade level, and knowledge points, which incorporates both outcome and process evaluations.
- GOAL: A Challenging Knowledge-grounded Video Captioning Benchmark for Real-time Soccer Commentary Generation46
Based on soccer game videos and synchronized commentary data, GOAL, a benchmark of over 8.9k soccer video clips, 22k sentences, and 42k knowledge triples is presented for proposing a challenging new task setting as Knowledge-grounded Video Captioning (KGVC).
- Joint Representation Learning of Cross-lingual Words and Entities via Attentive Distant Supervision46
This paper proposes a novel method for joint representation learning of cross-lingual words and entities that captures mutually complementary knowledge, and enables cross-lingsual inferences among knowledge bases and texts.
- 45
- Information Extraction45
This chapter will investigate the problems of information extraction and survey existing methodologies for solving these problems and several real-world applications of Information Extraction will be introduced.
- Image Enhanced Event Detection in News Articles44
This paper proposes a novel Dual Recurrent Multimodal Model, DRMM, to conduct deep interactions between images and sentences for modality features aggregation and superior performance compared to six state-of-art baselines as well as further ablation studies demonstrate the significance of image modality and effectiveness of the proposed architecture.
- 44
- ArnetMiner: an expertise oriented search system for web community44
An overview of the expertise oriented search system (ArnetMiner), in operation on the internet for more than one year, which indicates that users consider the system can really help people to find and share information in the web community.
- XLORE2: Large-scale Cross-lingual Knowledge Graph Construction and Application43
This work presents XLORE2, an extension of the XLORE that is built automatically from Wikipedia, Baidu Baike and Hudong Baike, and designs an entity linking system to demonstrate the effectiveness and broad coverage of XLore2.
- 43
- Panther43
This paper proposes a sampling method that provably and accurately estimates the similarity between vertices, based on a novel idea of random path, and shows that the proposed algorithm achieves clearly better performance than several alternative methods.
- CogCoM: A Visual Language Model with Chain-of-Manipulations Reasoning42
Chain of Manipulations is proposed, a mechanism that enables VLMs to solve problems with a series of manipulations, where each manipulation refers to an operation on the visual input, either from intrinsic abilities acquired through prior training or from imitating human-like behaviors.
- 42
- 42
- Interactive Contrastive Learning for Self-Supervised Entity Alignment41
This work conducts bidirectional contrastive learning via building pseudo-aligned entity pairs as pivots to achieve direct cross-KG information interaction and exploits the integration of entity textual and structural information and elaborately designs encoders for better utilization in the self-supervised setting.
- A gauss function based approach for unbalanced ontology matching41
This paper proposes a novel Gauss Function based ontology matching approach which outperforms the existing methods in terms of precision, recall and elapsed time and performs an extensive experiment to verify the effectiveness and efficiency of this approach.
- 40
- Toward Generalizable Evaluation in the LLM Era: A Survey Beyond Benchmarks39
This survey probes the core challenges that the rise of LLMs poses for evaluation, and identifies and analyzes two pivotal transitions: from task-specific to capability-based evaluation and from manual to automated evaluation.
- Large scale instance matching via multiple indexes and candidate selection38
This paper systematically investigates the characteristics of instance matching, and proposes a scalable and efficient instance matching approach named VMI, which generates multiple vectors for different kinds of intained in the ontology instances, and uses a set of inverted indexes based rules to get the primary matching candidates.
- EOS38
The experimental results demonstrate that the proposed methods for expert finding and association search in a social network are both more effective and efficient than the baseline methods.
- From MOOC to MAIC: Reshaping Online Teaching and Learning through LLM-driven Agents37
This work proposes MAIC (Massive AI-empowered Course), a new form of online education that leverages LLM-driven multi-agent systems to construct an AI-augmented classroom, balancing scalability with adaptivity.
- GraphQ IR: Unifying the Semantic Parsing of Graph Query Languages with One Intermediate Representation37
GraphQ IR has a natural-language-like expression that bridges the semantic gap and formally defined syntax that maintains the graph structure and can more precisely convert user queries into GraphQ IR, which can be later losslessly compiled into various downstream graph query languages.
- 37
- Mining competitive relationships by learning across heterogeneous networks37
The proposed model, Topical Factor Graph Model, defines a latent topic layer to bridge the two networks and learns a semi-supervised learning model to classify the relationships between entities (e.g., companies or products).
- StructInf: Mining Structural Influence from Social Streams35
This paper introduces a novel notion of structural influence and studies how to efficiently discover structural influence patterns from social streams by presenting three sampling algorithms with theoretical unbiased guarantee to speed up the discovery process.
- NewsMiner: Multifaceted news analysis for event search35
This article represents news as a link-centric heterogeneous network and formalize news analysis and mining task as link discovery problem and proposes a co-mention and context based knowledge linking method and a topic-level social content alignment method to establish the links between news and external sources.
- MoocRadar: A Fine-grained and Multi-aspect Knowledge Repository for Improving Cognitive Student Modeling in MOOCs34
This paper presents MoocRadar, a fine-grained, multi-aspect knowledge repository consisting of 2,513 exercise questions, 5,600 knowledge concepts, and over 12 million behavioral records, and proposes a framework to guarantee a high-quality and comprehensive annotation of fine- grained concepts and cognitive labels.
- Entity Matching across Heterogeneous Sources34
A probabilistic topic model is proposed that can not only find the matched documents for a query entity, but also explain why these documents are related by showing the common topics they share.
- Preference-aware Graph Attention Networks for Cross-Domain Recommendations with Collaborative Knowledge Graph33
A preference-aware Graph Attention network model with Collaborative Knowledge Graph (PGACKG) for cross-domain recommendations that aggregates the preference features of similar entities within domains and across domains to better learn user preference features is proposed.
- SWARMS33
This chapter describes the architecture and the main features of SWARMS, a platform for domain knowledge management that provides a suite of components for not only supporting efficient semantic data storage but also searching and mining the semantics.
- LLMAEL: Large Language Models are Good Context Augmenters for Entity Linking32
LLM-Augmented Entity Linking (LLMAEL) is introduced, the first framework to enhance specialized EL models with LLM data augmentation, which leverages off-the-shelf, tuning-free LLMs as context augmenters, generating entity descriptions to serve as additional input for specialized EL models.
- Are Missing Links Predictable? An Inferential Benchmark for Knowledge Graph Completion32
This work presents InferWiki, a Knowledge Graph Completion (KGC) dataset that improves upon existing benchmarks in inferential ability, assumptions, and patterns, and includes various inference patterns (e.g., reasoning path length and types) for comprehensive evaluation.
- Triple Classification Using Regions and Fine-Grained Entity Typing32
This work proposes a new region-based embedding approach using fine-grained type chains to extend the vectors of pre-trained entities into n-balls (n-dimensional balls) under the condition that head balls shall contain their tail balls.
- SQUIRE: A Sequence-to-sequence Framework for Multi-hop Knowledge Graph Reasoning31
SQUIRE is the first Sequence-to-sequence based multi-hop reasoning framework, which utilizes an encoder-decoder Transformer structure to translate the query to a path, and has the flexibility to complete missing edges along the path, especially in sparse KGs.
- DeepChannel: Salience Estimation by Contrastive Learning for Extractive Document Summarization31
DeepChannel is proposed, a robust, data-efficient, and interpretable neural model for extractive document summarization that reaches a ROUGE-1 F-1 score of 39.41 on CNN/Daily Mail test set with merely $1 / 100$ training set, demonstrating a tremendous data efficiency.
- Attributed and Predictive Entity Embedding for Fine-Grained Entity Typing in Knowledge Bases31
This work proposes an attributed and predictive entity embedding method, which can fully utilize various kinds of information comprehensively in knowledge bases and significantly outperforms 8 state-of-the-art methods.
- 31
- Pre-training language model incorporating domain-specific heterogeneous knowledge into a unified representation30
A heterogeneous knowledge language model (\textbf{HKLM}), a unified pre-trained language model (PLM) for all forms of text, including unstructured text, semi-structured text, and well-structured text, and well-structured text for multi-format text is proposed.
- Is Multi-Hop Reasoning Really Explainable? Towards Benchmarking Reasoning Interpretability28
A unified framework to quantitatively evaluate the interpretability of multi-hop reasoning models so as to advance their development and design an approximate strategy to calculate three metrics, including path recall, local interpretability, and global interpretability for evaluation, is proposed.
- Semantic Web and Web Science28
The book will focus on exploiting state of the art research in semantic web and web science and covering a number of global standards of the web and cutting edge technologies, such as: linked data, social semantic web, semanticweb search, smart data integration, semantic web mining and web scale computing.
- Building a large Chinese corpus annotated with semantic dependency28
This paper attempts to build a large corpus and annotate semantic knowledge with dependency grammar, and congruence is defined to measure the consistency of tagged corpus.
- StoryWriter: A Multi-Agent Framework for Long Story Generation26
This paper presents StoryWriter, a modular and open-source multi-agent framework for controllable and scalable long story generation, and significantly outperforms existing story generation baselines in both story quality and length.
- Generalized multipath planning model for ride-sharing systems26
A similarity model is provided which can reflect the personal preferences of the rides and utilize social media to obtain the current interests of the riders and drivers and allows each driver to generate sub-optimal paths according to his own requirements by suitably adjusting the weights.
- iASA: Learning to Annotate the Semantic Web26
iASA is a tool that learns to automatically annotate web documents according to an ontology that exploits machine learning methods to correctly select instances and to predict missing instances and shows that iASA can reach high accuracy quickly.
- 25
- Local dense mixed region cutting + global rebalancing: a method for imbalanced text sentiment classification25
A two-stage data balancing scheme for text sentiment classification, which not only can make the data boundary clear, but also can balance the class distribution of training data set.
- 24
- 24
- GLM-Dialog: Noise-tolerant Pre-training for Knowledge-grounded Dialogue Generation24
Comprehensive evaluations from automatic to human perspective demonstrate the advantages of GLM-Dialog comparing with existing open source Chinese dialogue models and offer the evaluation platform online in an effort to prompt the development of open source models and reliable dialogue evaluation systems.
- 24
- Fast and Flexible Top- k Similarity Search on Large Networks24
A sampling-based method using random paths to estimate the similarities based on both common neighbors and structural contexts efficiently in very large homogeneous or heterogeneous information networks.
- 24
- Role-aware conformity influence modeling and analysis in social networks24
This paper first formalizes conformity influence using a utility function based on the conformity theory from social psychology, and then proposes a probabilistic graphical model, referred to as Role-Conformity Model (RCM), for modeling the role-aware conformity influence between users by incorporating the utility function.
- A unified approach to matching semantic data on the Web24
A unified matching approach to finding equivalent entities in ontologies and LOD datasets on the Web that combines multiple lexical matching strategies using a novel voting-based aggregation method and utilizes the structural information and the already found correspondences to discover additional ones.
- Discovering the staring people from social networks24
A genetic algorithm based method is presented to solve the problem of staring people discovery from social networks, which is concerned with finding people who are not only authoritative but also sociable in the social network.
- 24
- Adapting Prime Number Labeling Scheme for Directed Acyclic Graphs24
The prime number labeling scheme is extended for DAG, which simplifies the transitive closure computations and diminishes storage redundancy, as well as inherits the dynamic labeling ability from original scheme.
- Weighted Ontology-Based Search Exploiting Semantic Similarity24
This paper first employs keyword based search method to retrieve instances for a query; then a proposed method of semantic feedback is performed to refine the search results; and then it conducts re-retrieval by making use of relations and instance similarities.
- XDAI: A Tuning-free Framework for Exploiting Pre-trained Language Models in Knowledge Grounded Dialogue Generation23
XDAI is proposed, a knowledge-grounded dialogue system that is equipped with the prompt-aware tuning-free PLM exploitation and supported by the ready-to-use open-domain external knowledge resources plus the easy- to-change domain-specific mechanism.
- 22
- 21
- LittleMu: Deploying an Online Virtual Teaching Assistant via Heterogeneous Sources Integration and Chain of Teach Prompts21
A virtual MOOC teaching assistant, LittleMu with minimum labeled training data, to provide question answering and chit-chat services on XuetangX MOOC platform, which continuously contributes to a more convenient and fair education.
- 21
- Explainable Few-Shot Knowledge Tracing20
This study proposes a cognition-guided framework that can track students’ knowledge from a few students’ records while providing natural language explanations, and results from three widely used datasets show that LLMs can perform comparable or superior to competitive deep KT methods.
- Plink-LDA: Using Link as Prior Information in Topic Modeling20
A prior information enabled topic model-PLDA that greatly outperforms basic LDA procedures on a clustering task while also maintaining the dependencies among documents and showing the feasibility by the task of citation recommendation.
- Shifting Long-Context LLMs Research from Input to Output19
This paper advocates for a paradigm shift in NLP research toward addressing the challenges of long-output generation and calls for focused efforts to develop foundational LLMs tailored for generating high-quality, long-form outputs, which hold immense potential for real-world applications.
- LC4EE: LLMs as Good Corrector for Event Extraction19
This work proposes an approach, L LMs-as-C orrector for E vent E xtraction ( LC4EE), aiming to leverage the superior extraction capability of SLMs and the instruction-following ability of LLMs to construct a robust and highly available EE sys-tem.
- 19
- Imposing Category Trees Onto Word-Embeddings Using A Geometric Construction19
A novel method to precisely impose tree-structured category information onto word-embeddings, resulting in ball embeddings in higher dimensional spaces, and experiments show that N -ball embeddings, carrying category information, outperform word embeddings in the test of nearest neighborhoods, and demonstrate surprisingly good performance in validating categories of unknown words.
- Transfer Learning Based Cross-lingual Knowledge Extraction for Wikipedia19
This paper forms the problem of cross-lingual knowledge extraction from multilingual Wikipedia sources, and presents a novel framework, called WikiCiKE, to solve this problem, and demonstrates that it outperforms the monolingualknowledge extraction method and the translation-based method.
- The dynamic features of Delicious, Flickr, and YouTube19
A Tagger Tag Resource Latent Dirichlet Allocation model (TTRLDA) is proposed to explore the evolution of topics emerging from social tagging vocabularies in Delicious, Flickr, and YouTube from 2003 to 2008.
- 18
- 18
- KQA Pro: A Large Diagnostic Dataset for Complex Question Answering over Knowledge Base.18
This work introduces KQA Pro, a large-scale dataset for Complex KBQA, and generates questions, SPARQLs, and functional programs with recursive templates and paraphrase the questions by crowdsourcing, giving rise to around 120K diverse instances.
- Ensemble method to joint inference for knowledge extraction18
An ensemble learning approach to joint inference on the three NLP sub-tasks is presented and it is shown that this approach leads to a higher precision and recall than that of those pipeline approaches.
- Conference Mining via Generalized Topic Modeling18
Experimental results show that proposed approach significantly outperformed baseline approach in discovering topically related conferences and finding conferences correlations because of its ability to produce less sparse topics.
- Topic-Level Random Walk through Probabilistic Model18
Experimental results on a real-world data set show that the proposed approach can significantly outperform the baseline methods of using language model and that of using traditional PageRank.
- Risk Minimization Based Ontology Mapping18
An approach called RiMOM to automatically discover mapping between ontologies, which explicitly and formally gives a complete decision model for ontology mapping and introduces a method to deal with instances heterogeneity, which is a long-standing problem for information processing.
- ReaRAG: Knowledge-guided Reasoning Enhances Factuality of Large Reasoning Models with Iterative Retrieval Augmented Generation17
This study enhances LRMs' factuality while effectively integrating robust reasoning for Retrieval-Augmented Generation (RAG) with ReaRAG, a factuality-enhanced reasoning model that explores diverse queries without excessive iterations.
- How do Transformers Learn Implicit Reasoning?17
This paper studies how implicit multi-hop reasoning emerges by training transformers from scratch in a controlled symbolic environment, revealing a three-stage developmental trajectory: early memorization, followed by in-distribution generalization, and eventually cross-distribution generalization.
- A Roadmap for Big Model17
This paper discusses not only the BM technologies themselves but also the prerequisites for BM training and applications with BMs, dividing the BM review into four parts: Resource, Models, Key Technologies and Application.
- XLink: An Unsupervised Bilingual Entity Linking System17
This paper builds an online bilingual entity linking system XLink, which is based on Wikipeida and Baidu Baike and proposes an unsupervised generative probabilistic method and utilize text and knowledge joint representations to perform entity disambiguation.
- A Generalized Topic Modeling Approach for Maven Search17
Experimental results on the corpus downloaded from DBLP show that proposed approach significantly outperformed the baseline approach, due to its ability to produce less sparse topics.
- Exploiting Temporal Authors Interests via Temporal-Author-Topic Modeling17
Experimental results on research papers dataset show the effectiveness of proposed approach and dominance over Author-Topic (AT) model, due to not changing the meaning of particular topic overtime.
- Academic conference homepage understanding using constrained hierarchical conditional random fields17
This paper proposes a unified approach, Constrained Hierarchical Conditional Random Fields, to accomplish the three labeling tasks simultaneously, and develops a prototype system of use-oriented semantic academic conference calendar.
- OpenEP: Open-Ended Future Event Prediction16
This article introduces OpenEP (an Open-Ended FEP task), which generates flexible and diverse predictions aligned with real-world scenarios, and proposes StkFEP, a stakeholder-enhanced FEP framework, that incorporates event characteristics for open-ended settings.
- 16
- RiMOM2013 results for OAEI 201316
The basic framework of the matching System (RiMOM 2013) is described and the alignment process and alignment strategies of RiMOM2013 are described, and specific techniques used for different tracks are presented.
- 16
- LinguaLens: Towards Interpreting Linguistic Mechanisms of Large Language Models via Sparse Auto-Encoder15
This work proposes LinguaLens, a systematic and comprehensive framework for analyzing the linguistic mechanisms of large language models, based on Sparse Auto-Encoders (SAEs), and extracts a broad set of Chinese and English linguistic features across four dimensions (morphology, syntax, semantics, and pragmatics).
- 15
- Discovering Missing Semantic Relations between Entities in Wikipedia15
This paper proposes an approach for automatically discovering the missing entity links in Wikipedia's infoboxes, and it significantly outperforms the baseline methods in terms of both precision and recall.
- A constraint-based topic modeling approach for name disambiguation15
A general framework of constraint-based topic modeling is proposed, which can make use of user-defined constraints to enhance the performance of name disambiguation.
- 15
- 15
- SoAy: A Solution-based LLM API-using Methodology for Academic Information Seeking14
SoAy enables LLMs to generate code for invoking APIs, guided by a pre-constructed API calling sequence referred to as a solution, which simplifies the model's understanding of complex API relationships, while the generated code enhances reasoning efficiency.
- AtomR: Atomic Operator-Empowered Large Language Models for Heterogeneous Knowledge Reasoning14
AtomR is introduced, a framework for LLMs to conduct accurate heterogeneous knowledge reasoning at the atomic level and three atomic knowledge operators are proposed, a unified set of operators for LLMs to retrieve and manipulate knowledge from heterogeneous sources.
- GEOSatDB: global civil earth observation satellite semantic database14
The development of GEOSatDB is described, the most complete semantic database of civil Earth observation satellites developed based on a unified ontology model, which contains 127,949 semantic statements for 2,340 remote sensing satellites and 1,021 observation sensors.
- 14
- 14
- Interpretable and Low-Resource Entity Matching via Decoupling Feature Learning from Decision Making14
A novel EM framework that consists of Heterogeneous Information Fusion (HIF) and Key Attribute Tree (KAT) Induction to decouple feature representation from matching decision and is highly efficient and outperforms SOTA EM models in most cases.
- Cooperative Hybrid Semi-Supervised Learning for Text Sentiment Classification14
A synthetical framework that covers the whole process of semi-supervised learning from seed selection, iterative modification of the training text set, to the co-training strategy of the classifier is proposed in this paper for text sentiment classification.
- Name List Only? Target Entity Disambiguation in Short Texts14
This paper proposes a graphbased model called TremenRank to collectively identify target entities in short texts given a name list only that propagates trust within the graph, allowing for an arbitrary number of target entities and texts using inverted index technology.
- Role-Aware Conformity Modeling and Analysis in Social Networks14
This paper formalizes conformity using a utility function based on the conformity theory from social psychology, and extends and incorporates the utility function into a probabilistic topic model, called the Role-Conformity Model (RCM), for modeling user behaviors under the effect of conformity.
- Cross-Lingual Knowledge Validation Based Taxonomy Derivation from Heterogeneous Online Wikis14
This paper investigates different linguistic heuristics and language independent features, and proposes a cross-lingual knowledge validation based dynamic adaptive boosting model to iteratively reinforce the performance of taxonomic relation prediction.
- Incorporating Entities in News Topic Modeling14
A generative model, Entity-Centered Topic Model (ECTM) is proposed to summarize the correlation among entities, words and topics by taking entity topic as a mixture of word topics.
- 13
- 13
- TWAG: A Topic-Guided Wikipedia Abstract Generator13
A two-stage model TWAG is proposed that guides the abstract generation with topical information and outperforms various existing baselines and is capable of generating comprehensive abstracts.
- Learning to Infer Competitive Relationships in Heterogeneous Networks13
This article uses Twitter and patent records as data sources and statistically study the patterns behind the competitive relationships and proposes a novel Topical Factor Graph Model, which employs a semi-supervised learning algorithm to classify the relationships between entities.
- Forecasting Potential Diabetes Complications13
A novel probabilistic model called Sparse Factor Graph Model (SparseFGM), which projects sparse features onto a lower-dimensional latent space, which alleviates the problem of sparseness and is able to capture the associations between complications and lab tests.
- Labeling clusters from both linguistic and statistical perspectives: A hybrid approach13
Experimental results show that the proposed approach produces high quality labels and outperforms existing state-of-art methods on both manual and automatic evaluations.
- Loss Minimization Based Keyword Distillation13
A new approach based on Bayesian decision theory is presented, taking the word’ context and linkages between words into account, and it is shown that this approach significantly improves the quality of extracted keywords.
- ViLTA: Enhancing Vision-Language Pre-training through Textual Augmentation12
A novel method ViLTA is proposed, comprising of two components to further facilitate the model to learn fine-grained representations among image-text pairs, and can achieve better performance on various vision-language tasks.
- 12
- 12
- Arnetminer: expertise oriented search using social networks12
The results demonstrate that the proposed relevancy propagation expert finding method outperforms the method that only uses person local information, and the proposed two-stage association search on a large-scale social network is order of magnitude faster than the baseline method.
- JiuZhou: open foundation language models and effective pre-training framework for geoscience11
This study constructs the large-scale, diverse, and high-quality JiuZhou-Corpus and the JiuZhou-Framework specifically designed for training geoscience large language models (LLMs), and introduces a two-stage pre-adaptation pre-training method to enhance the efficiency of knowledge learning and transfer in the model.
- 11
- Exploring Universal Intrinsic Task Subspace for Few-Shot Learning via Prompt Tuning11
Evidence is empirically found indicating that the adaptations of PLMs to various few-shot tasks can be reparameterized as optimizing only a few free parameters in a unified low-dimensional intrinsic task subspace, which may help to understand why PLMs could easily adapt to various NLP tasks with small-scale data.
- Evaluating Generative Language Models in Information Extraction as Subjective Question Correction11
Results on three information extraction tasks show that SQC-Score is more preferred by human annotators than the baseline metrics, and by incorporating a Natural Language Inference model, SQC-Score enriches golden labels, addressing benchmark incompleteness by acknowledging correct yet previously omitted answers.
- EDUKG: a Heterogeneous Sustainable K-12 Educational Knowledge Graph11
This work designs an interdisciplinary and fine-grained ontology for uniformly modeling knowledge and resource in K-12 education, and proposes a flexible methodology for interactively extracting factual knowledge from textbooks called EDUKG, a heterogeneous sustainable K- 12 Educational Knowledge Graph.
- 11
- Learning to Align Comments to News Topics11
This article formally defines the news comment alignment problem and proposes a novel framework that automatically extracts topics from a given news article and its associated comments and completes the alignment between news sentences and comments through a weighted-SVM classifier.
- Ride-Sharing: A Multi Source-Destination Path Planning Approach11
This paper formalized the ride sharing problem as a multi source-destination path planning problem, and developed an objective function which models different conflicting objectives in a unified framework and has the potential to compute multiple optimal paths.
- Building a Large Scale Knowledge Base from Chinese Wiki Encyclopedia11
This paper aims at building large scale Chinese structured knowledge base from Hudong, which is one of the largest Chinese Wiki Encyclopedia websites and provides RDF dump or SPARQL to access the established Chinese knowledge base.
- RiMOM2: A Flexible Ontology Matching Framework11
This paper presents a tool RiMOM2 which provides a framework for both ontology schema matching and instance matching, and aims to solve the schema heterogeneity problem in the Linking Open Data project.
- Constraint Back-translation Improves Complex Instruction Following of Large Language Models10
This work finds that existing datasets inherently contain implicit complex constraints and proposes a novel data generation technique, constraint back-translation, which improves multiple backbone LLMs' complex instruction-following ability, evaluated on extensive instruction-following benchmarks.
- How Proficient Are Large Language Models in Formal Languages? An In-Depth Insight for Knowledge Base Question Answering10
Evaluating the understanding and generation ability of LLMs to deal with differently structured logical forms by examining the inter-conversion of natural and formal language through in-context learning of LLMs indicates that LLMs exhibit considerable sensitivity.
- Unveiling the Black Box of PLMs with Semantic Anchors: Towards Interpretable Neural Semantic Parsing10
This work proposes to incorporate the current PLMs with a hierarchical decoder network, taking the first-principle structures as the semantic anchors, and proposes two novel intermediate supervision tasks for training the hierarchical decoders and probing the model intermediate representations in a self-adaptive manner alongside the fine-tuning process.
- Distantly Supervised Course Concept Extraction in MOOCs with Academic Discipline10
A novel three-stage framework DS-MOCE is presented, which leverages the power of pre-trained language models explicitly and implicitly and employs discipline-embedding models with a self-train strategy based on label generation refinement across different domains and provides an expert-labeled dataset spanning 20 academic disciplines.
- 10
- Construction of Multimodal Chinese Tourism Knowledge Graph10
A tourism route generation algorithm is proposed, which could automatically schedule the tourism routes by incorporating the characteristics of attraction and tourists’ demands and a sharing platform is established, which provides open data access and query interface.
- ExpanRL: Hierarchical Reinforcement Learning for Course Concept Expansion in MOOCs10
Employing a two-level HRL mechanism of seed selection and concept expansion, ExpanRL is more feasible to adjust the expansion strategy to find new concepts based on the students’ feedback on expansion results.
- Domain-Specific Entity Linking via Fake Named Entity Detection10
This paper proposes a novel approach to jointly model NED and EL for domain-specific knowledge base, facilitating mentions extracted from unstructured data to be accurately matched to uniquely identifiable entities in the given domain- specific knowledge base.
- Learning Topic Hierarchies for Wikipedia Categories10
This paper aims to explore multi-level topic knowledge contained in Wikipedia articles by utilizing Wikipedia for various knowledge acquisition tasks and exploring the role of Wikipedia articles in this regard.
- 10
- 10
- 10
- Recommendation based on object typicality10
This paper investigates recommendation systems from a new perspective based on object typicality and proposes a novel typicality-based recommendation approach that outperforms compared methods on recommendation quality.
- Query-Focused Summarization by Combining Topic Model and Affinity Propagation10
This paper proposes a novel approach by combining statistical topic model and affinity propagation, called qLDA, that can simultaneously model documents and the query and can automatically discover key sentences from the document collection without predefining the length of the summary.
- 10
- A Semantic Matchmaker for Ranking Web Services10
A new method of semantic matchmaker is proposed to use the semantic distance for estimating the matching degree between a service and a user request.
- 10
- R-Eval: A Unified Toolkit for Evaluating Domain Knowledge of Retrieval Augmented Large Language Models9
The R-Eval toolkit is introduced, a Python toolkit designed to streamline the evaluation of different RAG workflows in conjunction with LLMs and emphasizes the importance of considering both task and domain requirements when choosing a RAG workflow and LLM combination.
- 9
- Reverse That Number! Decoding Order Matters in Arithmetic Learning9
This work introduces a novel strategy that not only reevaluates the digit order by prioritizing output from the least significant digit but also incorporates a step-by-step methodology to substantially reduce complexity.
- 9
- 9
- 9
- 9
- 9
- 9
- 9
- Event-level Knowledge Editing8
A new task setting is proposed: event-level knowledge editing, which directly edits new events into LLMs and improves over conventional triplet-level editing on efficiency and completeness.
- Towards a General Pre-training Framework for Adaptive Learning in MOOCs8
Through a series of downstream tasks of Learning Recommendation, Learning Resource Evaluation, Knowledge Tracing, and Dropout Prediction, it is found that course structures, text, and knowledge are helpful for modeling and inherently coherent to student non-sequential learning behaviors and that indirectly relevant information included in the pre-training foundation can be shared across downstream tasks to facilitate effectiveness.
- Category Multi-representation: A Unified Solution for Named Entity Recognition in Clinical Texts8
This paper proposes a novel method that learn multiple representations for each category, namely category-multi-representation (CMR) that captures the semantic relatedness between words and clinical categories from different perspectives and enjoys a lightweight adaption among languages.
- 8
- What users care about: a framework for social content alignment8
A novel framework is proposed that is able to automatically detect the sub topics from a given Web document, and also align the associated social comments with the detected subtopics and significantly outperforms the existing state-of-the-art methods in social content alignment.
- A unified framework for name disambiguation8
The name disambiguation problem is formalized in a unified framework that can incorporate both attribute and relationship into a probabilistic model and Experimental results show that the proposed framework can significantly outperform the baseline method.
- 8
- GeoFactory: an LLM performance enhancement framework for geoscience factual and inferential tasks7
High-level guidance for the selection and application of LLM enhancement methods in geoscience is offered and it is shown that, through appropriate LLM enhancement, a model with fewer parameters can achieve a performance comparable to that of GPT-4 in geoscience tasks.
- Learning in Context: Personalizing Educational Content with Large Language Models to Enhance Student Learning7
This work introduces PAGE, a novel framework that leverages large language models (LLMs) to automatically personalize educational materials by adapting them to each student's unique context, such as their major and personal interests, and demonstrates the practical value of LLM-powered personalization.
- A Cause-Effect Look at Alleviating Hallucination of Knowledge-grounded Dialogue Generation7
Based on the causal effect analysis, a possible solution for alleviating the hallucination in KGD by exploiting the dialogue-knowledge interaction is proposed and results show this method can reduce hallucination without disrupting other dialogue performance, while keeping adaptive to different generation models.
- 7
- 7
- 7
- 7
- 7
- 7
- Neural Gibbs Sampling for Joint Event Argument Extraction7
A Bayesian model named Neural Gibbs Sampling (NGS) is proposed to jointly extract event arguments by training two neural networks to model the prior distribution and conditional distribution over event arguments respectively and then using Gibbs sampling to approximate the joint distribution with the learned distributions.
- Cross-Lingual Infobox Alignment in Wikipedia Using Entity-Attribute Factor Graph7
This paper proposes entity-attribute factor graph to consider not only individual features but also the correlations among attribute pairs, and proposes 8 effective features for cross lingual infobox attribute matching containing categories, templates, attribute labels and values.
- Proceedings of the 2nd ACM workshop on Social web search and mining7
The call for papers attracted 17 submissions from Asia, Europe, and the United States and the program committee accepted 8 full papers and 1 short paper that cover a variety of topics, including social network modeling and analysis, social influence analysis, influenceAnalysis, and semantic social network.
- SWSCF: A Semantic-based Web Service Composition Framework7
This paper presents a semantic-based Web service composition framework called SWSCF based on both semantic process recommendation and semantic matchmaking and proves it to be more flexible and effective than some service composition frameworks.
- Semantic Similarity Based Ontology Cache7
Experimental results indicate that the proposed methods can significantly outperform the baseline methods for ontology caching and have been applied to a research project that is called SWARMS.
- Association search in semantic web: search + inference7
This paper first minutely defines the association search and its categorization, then takes ontology taxonomy as network structure in Bayesian network and gives the solution for association search in the network.
- 6
- WebSeer: Training Deeper Search Agents through Reinforcement Learning with Self-Reflection6
This paper constructs a large dataset annotated with reflection patterns and design a two-stage training framework that unifies cold start and reinforcement learning within the self-reflection paradigm for real-world web-based environments, which enables the model to generate longer and more reflective tool-use trajectories.
- LongWriter-V: Enabling Ultra-Long and High-Fidelity Generation in Vision-Language Models6
This work introduces LongWriter-V-22k, a SFT dataset comprising 22,158 examples, each with multiple input images, an instruction, and corresponding outputs ranging from 0 to 10,000 words, and proposes IterDPO, which breaks long outputs into segments and uses iterative corrections to form preference pairs with the original outputs.
- 6
- 6
- KACC: A Multi-task Benchmark for Knowledge Abstraction, Concretization and Completion6
This paper proposes large-scale datasets extracted from Wikidata, which provide more size-balanced concept graphs and abundant cross-view links and proposes a benchmark to test the ability of existing models on knowledge abstraction, concretization and completion (KACC).
- 6
- Overview of CCKS 2018 Task 1: Named Entity Recognition in Chinese Electronic Medical Records6
The CCKS 2018 presented a named entity recognition (NER) task focusing on Chinese electronic medical records (EMR), and showed that the traditional CRF and Bi-LSTM model were the most popular models for the task.
- Type Hierarchy Enhanced Heterogeneous Network Embedding for Fine-Grained Entity Typing in Knowledge Bases6
A novel embedding based hierarchical entity typing framework which uses learning to rank algorithm to enhance the performance of word-entity-type network embedding and can take full advantage of labeled and unlabeled data.
- o-HETM: An Online Hierarchical Entity Topic Model for News Streams6
A Hierarchical Entity Topic Model (HETM) is proposed which considers the timeliness of news data and the importance of named entities in conveying information of who/when/where in news articles and greatly improves the topic quality and time efficiency.
- TSDPMM: Incorporating Prior Topic Knowledge into Dirichlet Process Mixture Models for Text Clustering6
Experimental results on document clustering across three datasets demonstrate the proposed TSDPMM significantly outperforms state-of-the-art DPMM model and can be applied in a lifelong learning framework.
- Semantic Mining of Social Networks6
This book presents the architecture of the research for social network mining, from a microscopic point of view, and uses an academic social network, ArnetMiner, as an example to demonstrate how the introduced technologies for mining real social networks are applied.
- RiMOM-IM results for OAEI 20146
The overall framework of the matching System (RiMOM-IM) is described; the techniques used in the framework for instance matching are detailed and a thorough analysis on the results is given.
- Entire solution path for support vector machine for positive and unlabeled classification6
An algorithm is given for the entire solution path by choosing the ‘best’ cost parameter while training the SVM model and the results show that the algorithm achieves better results when dealing with positive data and unlabeled classification.
- A New Approach to Personal Network Search based on Information Extraction6
A new approach based on information extraction for constructing a personal network by discovering and identifying personal information from the web by finding and identifyingPersonal Network Search.
- 6
- A Model for Word Sense Disambiguation6
A model for mapping a structural semantic space from a thesaurus into a multi-dimensional, real-valued vector space is put forward and a word sense disambiguation method based on this mapping is given.
- An Improved Maximum Entropy Language Model and Its Application6
An improved maximum entropy approach which makes use of mutual information of information theory to select features based on Ztest is proposed and is applied to Chinese word sense disambiguation.
- Awaking the Slides: A Tuning-free and Knowledge-regulated AI Tutoring System via Language Model Coordination5
Slide2Lecture is developed, a tuning-free and knowledge-regulated intelligent tutoring system that can effectively convert an input lecture slide into a structured teaching agenda consisting of a set of heterogeneous teaching actions.
- 5
- TacoERE: Cluster-aware Compression for Event Relation Extraction5
A cluster-aware compression method for improving event relation extraction (TacoERE) is proposed, which explores a compression-then-extraction paradigm and introduces document clustering for modeling event dependencies.
- Mastering the Task of Open Information Extraction with Large Language Models and Consistent Reasoning Environment5
This paper proposes a method to effectively estimate the discrepancy of syntactic distribution between a LLM and test samples, which can serve as correlation evidence for preparing positive demonstrations and introduces a simple yet effective mechanism to establish the reasoning environment for LLMs on specific tasks.
- 5
- 5
- 5
- Metadata Extraction for Scientific Papers5
This paper presents a supervised method with three-stages to address the problem of metadata extraction for scientific literature, and achieves a +3.2% improvement in accuracy compared with strong baseline methods.
- Incremental learning from news events5
This paper formally defines the problem of incremental knowledge learning from similar news events on the same subject, where each event consists of a set of news articles reporting about it, and proposes a scalable two-phase framework to incrementally learn a topic hierarchy for a subject from Events on the subject as the events occur.
- A Semantics Enabled Intelligent Semi-structured Document Processor5
In semantic view extraction, by using similarity based rule induction, the approach can significantly outperform most of the existing wrapper methods and make use of the semantics that resides in document logical structure to help find relations between semantic entities.
- 5
- 5
- System II5
This paper proposes a native RDF repository that takes the hypergraph representation for RDF as the data model for its persistent storage, which effectively avoids the costs of data model transformation when accessing RDF data.
- 5
- Feature-Correlation Based Multi-view Detection5
A general algorithm CCFP (Correlation and Compatibility based Feature Partitioner) is presented to automate multi-view detection and the experiment results show that multi- view learner with views generated by CCFP outperforms learners with viewsgenerated by other means clearly.
- A Component Retrieval Method Based on Facet-Weight Self-learning5
In this paper, an intelligent component retrieval model – FWRM is proposed and the experimental results prove that this method is feasible and can improve component retrieval effectively.
- Personalized Learning Path Planning through Goal-Driven Learner State Modeling4
Pxplore is introduced, a novel framework for PLPP that integrates a reinforcement-based training paradigm and an LLM-driven educational architecture that designs a structured learner state model and an automated reward function that transforms abstract objectives into computable signals.
- EduCraft: A System for Generating Pedagogical Lecture Scripts from Long-Context Multimodal Presentations4
EduCraft is introduced, a novel system designed to automate Lecture Script Generation (LSG), addressing key difficulties such as comprehensive multimodal understanding, long-context coherence, and instructional design efficacy.
- 4
- 4
- Let Me Show You Step by Step: An Interpretable Graph Routing Network for Knowledge-based Visual Question Answering4
A novel interpretable graph routing network (GRN) which explicitly conducts entity routing over a constructed scene knowledge graph step by step for KB-VQA and clearly provides the reasoning process of KB-VQA and can handle the questions with unseen answers without distinction.
- Aligning Teacher with Student Preferences for Tailored Training Data Generation4
This work proposes ARTE, dubbed Aligning TeacheR with StudenT PreferencEs, a framework that aligns the teacher model with student preferences to generate tailored training examples for Knowledge Distillation, and thoroughly investigates the generalization of ARTE.
- Untangle the KNOT: Interweaving Conflicting Knowledge and Reasoning Skills in Large Language Models4
A new dataset is constructed, dubbed KNOT, for knowledge conflict resolution examination in the form of question answering, which facilitates in-depth analysis by dividing reasoning with conflicting knowledge into three levels: Direct Extraction, which directly extracts conflicting knowledge to answer questions, Explicit Reasoning, which reasons with conflicting knowledge when the reasoning path is explicitly provided in the question, and Implicit Reasoning, where reasoning with conflicting knowledge requires LLMs to infer the reasoning path independently to answer questions.
- How Can Cross-lingual Knowledge Contribute Better to Fine-Grained Entity Typing?4
By utilizing multilingual transfer learning via the mixture-of-experts approach, the model dynamically capture the relationship between target language and each source language, and effectively generalize to predict types of unseen entities in new languages.
- 4
- Geographical Information Enhanced POI Hierarchical Classification4
This paper proposes an Ensemble POI Hierarchical Classification framework (EHC) consisting of three components: Textual and Geographic Feature Extraction, Hierarchy Classifier, and Soft Voting Ensemble Model.
- Domain Adaptive Question Answering over Knowledge Base4
A domain adaptive method to construct a domain-specific question answering system using easily accessible open domain questions using generalization features to represent questions, which can categorize questions according to their syntactic forms.
- 4
- On Modeling Sense Relatedness in Multi-prototype Word Embedding4
A novel approach to capture word sense relatedness in multi-prototype word embedding model by introducing a random process to integrate these two types of senses and design two non-parametric methods for word sense induction.
- 4
- The design of a live social observatory system4
NExT-Live offers a live observatory platform that enables people to know the happenings of the place in order to lead better life and generates high-level analytics about the sense of different target entitles, including People, Locations, Topics and Organizations.
- Exploration and Visualization of Administrator Network in Wikipedia4
An administrator exploration prototype system which can rank the selected characteristics for administrators and can be used as a decision support system is developed and some meaningful observations are found to show that the administrator network is a healthy small world community.
- A Composition Oriented and Graph-Based Service Search Method4
A composition-oriented and AND/OR graph-based service search method named CoSA, which can automatically search the composition relation graph in concepts level and composition plan in implement level and enables the dynamic and automatic service composition.
- 4
- Transforming Heterogeneous Messages Automatically in Web Service Composition4
A new solution for linking heterogeneous messages automatically when composing web services that converts request messages to the format of current service’s input and is applied in SEWSIP, a prototype of semantic-based web service composition system and evaluation of related experiments shows well results.
- 4
- 4
- 4
- Knowledge-to-Jailbreak: Investigating Knowledge-driven Jailbreaking Attacks for Large Language Models3
A new task, knowledge-to-jailbreak, is proposed, which aims to generate jailbreaking attacks from domain knowledge, requiring both attack effectiveness and knowledge relevance, and shows that jailbreak-generator can generate jailbreaks that are comparable in harmfulness to those crafted by human experts.
- SIRI: Scaling Iterative Reinforcement Learning with Interleaved Compression3
SIRI, Scaling Iterative Reinforcement Learning with Interleaved Compression, a simple yet effective RL approach for Large Reasoning Models that enables more efficient and accurate reasoning by dynamically adjusting the maximum rollout length during training is introduced.
- 3
- LecEval: An Automated Metric for Multimodal Knowledge Acquisition in Multimedia Learning3
LecEval, an automated metric grounded in Mayer's Cognitive Theory of Multimedia Learning, is introduced to evaluate multimodal knowledge acquisition in slide-based learning, bridging the gap between automated and human assessments.
- 3
- 3
- 3
- 3
- 3
- 3
- HOSMEL: A Hot-Swappable Modularized Entity Linking Toolkit for Chinese3
This work investigates the usage of entity linking in downstream tasks and presents the first modularized EL toolkit for easy task adaptation, HOSMEL, for Chinese, with three flexible usage modes, a live demo, and a demonstration video.
- 3
- Learning Multilingual Sentence Embeddings from Monolingual Corpus3
This work proposes a novel self-learning approach which is capable of learning multi-lingual sentence embeddings from monolingual corpora and outperforms existing unsupervised approaches and is competitive to supervised approaches.
- Learning to Embed Sentences Using Attentive Recursive Trees3
An end-to-end reinforced training strategy for AR-Tree is proposed, which is demonstrated to consistently outperform, or be at least comparable to, the state-of-the-art sentence embedding methods on three sentence understanding tasks.
- Domain Specific Cross-Lingual Knowledge Linking Based on Similarity Flooding3
This paper proposes an unsupervised framework for cross-lingual knowledge linking on specific domains, different from traditional methods, and evaluates the proposed method on two knowledge linking tasks to find English-Chinese knowledge links.
- RiMOM results for OAEI 2016.3
The overall framework of RiMOM (RiMOM), which is designed for large-scale and cross-lingual instance matching task specially, is described and an extended version, which provides support for cross-lingual instance matching in a supervised or an unsupervised way is proposed.
- RiMOM Results for OAEI 20153
The overall framework of the matching framework for instance matching (RiMOM) is described and the techniques used in the framework for instance matching for instance matching are detailed.
- On Modelling Non-linear Topical Dependencies3
An efficient expectation-maximization (EM) procedure for parameter estimation is developed and results show that GTRF achieves much lower perplexity than LDA and linear dependency topic models and produces better topic coherence.
- 3
- 3
- 3
- Semantic Annotation using Horizontal and Vertical Contexts3
Experimental results indicate that the proposed two-stage method for semantic annotation using machine learning can significantly outperform the baseline method as well as the sequence-based method for Semantic annotation.
- Term Committee Based Event Identification within News Topics3
A new event identification method based on the term committee is proposed, which first captures some tight term clusters as term committees of potential events, and then uses them to reweight the key terms in a story.
- Managing Large Scale Native RDF Semantic Repository from the Graph Model Perspective3
A native storage instead of relational database is used to hold RDF and indices supporting regular path expres- sion, full-text retrieval and partial OWL Lite inference are built above the storage model.
- 3
- Logical structure based semantic relationship extraction from semi-structured documents3
This paper proposes a method for semantic relationship extraction by using the logical information in the semi-structured document by using the logical information in the semi-structured document.
- 3
- Table Detection from Plain Text Using Machine Learning and Document Structure3
Experimental results indicate that the proposed methods can significantly outperform the baseline methods for unformatted table extraction and use document structure to improve the detection performance.
- MMGeoLM: Hard Negative Contrastive Learning for Fine-Grained Geometric Understanding in Large Multimodal Models2
This work proposes a novel hard negative contrastive learning framework for the vision encoder, which combines image-based contrastive learning using generation-based hard negatives created by perturbing diagram generation code, and text-based contrastive learning using rule-based negatives derived from modified geometric descriptions and retrieval-based negatives selected based on caption similarity.
- 2
- 2
- 2
- 2
- LM-Interview: An Easy-to-use Smart Interviewer System via Knowledge-guided Language Model Exploitation2
LM-Interview is introduced, a novel system designed to automate the process of preparing, conducting and analyzing semi-structured interviews and demonstrates that LM-interview achieves performance comparable to that of skilled human interviewers.
- 2
- 2
- 2
- 2
- ConstGCN: Constrained Transmission-based Graph Convolutional Networks for Document-level Relation Extraction2
A novel graph convolutional network which performs knowledge-based information propagation between entities along with all specific relation spaces without any prior graph construction, and constrain the propagation using transmitting scores learned from the Noise Contrastive Estimation between fact triples.
- 2
- 2
- 2
- Event-Oriented Wiki Document Generation2
The proposed WikiGen model, which consists of two parts, is capable of generating fluent and comprehensive Wikipedia documents and outperforms previous work, achieving state-of-the-art ROUGE scores.
- How Important Is POS to Dependency Parsing? Joint POS Tagging and Dependency Parsing Neural Networks2
This work proposes a method to jointly learn POS tagging and dependency parsing so as to alleviate the error propagation problems and reports the results support the idea that POS tagging is a crucial syntactic component for dependency parsing.
- Leveraging Multi-head Attention Mechanism to Improve Event Detection2
The proposed MANN is a multi-head attention mechanism model enhanced by argument knowledge that gives MANN the ability to detect a variety of information in a sentence while argument knowledge acts as a supervisor to further improve the quality of attention.
- 2
- How do pronouns affect word embedding2
Experiments show that by using co-reference resolution, the word embedding performance in the word analogy task can be improved by around 1.88 %, and that the words that are names of countries are affected the most, which is as expected.
- Detecting Public Influence on News Using Topic-Aware Dynamic Granger Test2
A novel topic-aware dynamic Granger test framework is proposed to quantify and characterize the public influence on news, and the results show promising prospects on predicting whether an event will be properly handled at its early stage.
- Graph-Based Jointly Modeling Entity Detection and Linking in Domain-Specific Area2
An iterative graph-based algorithm to jointly model the ED and EL tasks in domain-specific area by capturing the local dependency of mention- to-entity and the global interdependency of entity-to-entity is proposed.
- Measuring the Influence from User-Generated Content to News via Cross-dependence Topic Modeling2
This paper proposes a cross-dependence temporal topic model (CDTTM) for topic extraction, employs a hybrid method to discover short and long term influence links across streams, and introduces four measures to quantify how the unique topics from one stream affect or influence the generation of the other stream.
- 1 Toward integration travel information data using information extraction and instance matching2
The concept of instance matching is extended to find the connotative relationship between instances extracted from different sources in order to improve the result of integration of travel data with comprehensive information.
- NEI2
A novel four-phase framework NEI(News Event Insight) that focuses on visualizing a news event properly and clearly and demonstrates the quantifying properties of topics related to a certain event.
- Introduction to the Special Issue on Linking Social Granularity and Functions2
This special issue covers decreasing the granularity of network structure features for solving micro-level problems like user interest profiling and link sign prediction or improving the efficiency of time-consuming information maximization, moving beyond network clusters themselves by studying their overlap regions, which paves a contradictive way against conventional community detection, and leveraging both macro-levelnetwork structure features and micro- level node attributes for link prediction and community detection.
- Community Based User Behavior Analysis on Daily Mobile Internet Usage2
This paper proposes a framework which can calculate the proper number of the clusters in mobile user network and uses communities/clusters to work out the various interests of the users on the timeline of the day.
- Evaluating Article Quality and Editor Reputation in Wikipedia2
This paper proposes a dual wing factor graph (DWFG) model, which utilizes the mutual reinforcement between articles and editors to generate article quality and editor reputation and designs an efficient algorithm to learn the proposed factor graph model.
- 2
- Research on Mapping OWL Ontology to Software Code Model2
The method to map OWL ontology to AspectJ aspect-oriented model is proposed, which implements the mapping to software code model well and points out some problems existed in mapping directly to Java object- oriented model.
- Discovery and Query: Two Semantic Processes for Web Services2
The whole architecture for discovery and query is proposed, then four algorithms to discover web services and three algorithms to query web service based on the service data instance concept and similarity calculation are given.
- Rule-based Semantic Object Matching Algorithm2
The paper introduces the algorithm, which converts rules to DFA and analyzes the input words that match the DFA's transitions continuously, and then computes the similarity between the input text and the rule.
- 2
- Keyword Extraction Based Peer Clustering2
This paper proposes a method for clustering peers using the exchange text documents between them based on the peers' similarities and based on Bayesian decision theory.
- WODOS:Research and Implementation of the Middleware for Semantic Web2
The WODOS (Web Oriented Distributed Ontology System) is brought forward, which is a new middleware of the distributed ontology system in Web environment that can store, refer and process the ontology distributively and ensure the consistency and uniformity of the ontological system in web environment.
- 1
- VocQuiz: Vocabulary Question Generation for English Language Education1
VocQuiz is presented, a vocabulary question generation system that combines generalization capabilities of LLMs with reliable language resources, including dictionaries, NLP datasets and authentic corpora, to enhance both contextual relevance and linguistic accuracy.
- 1
- 1
- DiaKoP: Dialogue-based Knowledge-oriented Programming for Neural-symbolic Knowledge Base Question Answering1
DiaKoP enables users to decompose complex questions into multiple simpler follow-up questions and interact with the system to obtain answers, and implements a human-in-the-loop mechanism, allowing users to correct errors.
- 1
- 1
- Step out of KG: Knowledge Graph Completion via Knowledgeable Retrieval and Reading Comprehension1
A new model based on information retrieval and reading comprehension, namely IR4KGC is proposed, which can well solve relations that cannot be inferred from existing knowledge, and achieve good results on KGC datasets.
- CStory: A Chinese Large-scale News Storyline Dataset1
This work proposes a sophisticated pre-processing method to filter candidate news pairs by entity co-occurrence and semantic similarity and constructs CStory, a large-scale Chinese news storyline dataset, which contains 11,978 news articles, 112,549 manually labeled storyline relation pairs, and 49,832 evidence sentences for annotation judgment.
- 1
- 1
- 1
- Knowledge Fragment Enrichment Using Domain Knowledge Base1
The proposed model is able to model the dependencies among concepts in the input knowledge fragment and also capture the probabilistic relationship between concepts and domain entities.
- Boosting to Build a Large-Scale Cross-Lingual Ontology1
This paper proposes a cross-lingually boosting approach to iteratively reinforce the performance of ontology building and instance matching, and outputs an ontology containing over 3,520,000 English instances, 800,000 Chinese instances, and over 150,000 cross-lingsual instance alignments.
- 1
- 1
- Cross-Lingual Cross-Media Content Linking: Annotations and Joint Representations (Dagstuhl Seminar 15201)1
Dagstuhl Seminar 15201 was conducted on "Cross-Lingual Cross-Media Content Linking: Annotations and Joint Representations".
- Personalized Model Combination for News Recommendation in Microblogs1
In this model, the first level is defining four atomic recommendation models that make fully use of the social and content information of users and the second level is adapting to each user that atomic models effectively used.
- 1
- Inferring Public and Private Topics for Similar Events1
A topic model to infer the public and private topic from a group of similar events by using Bernoulli distribution to determine the source of words is proposed, aiming at the consistency and mapping of topics.
- Toward Ontology Representation and Reasoning for News1
An ontology-enriched news deep understanding framework ONDU is proposed which addresses the problems of how to transfer parsed news content into logical triples by using domain ontology and how to detect the inconsistence among multiple news articles and compare the different information implied in different news.
- Co-mention and Context-Based Entity Linking1
The approach first recognizes three types of named entities (i.e., person, location, and organization) and then uses a disambiguation method to link entities occurring in news articles to entities in knowledge bases.
- Personalized Diversity Search Based on User’s Social Relationships1
A systematic method named personalized diversity search based on user’s social relationships (PDSSR), this method is a combination of personalization and diversification, which enables computer better understand user�'s search intents and interests, consequently returns a personalized and reduced diversified result set.
- A Mathematical Model for Human Flesh Search Engine1
Typical Human Flesh Search Engine cases and simulation-based experiments are used to evaluate the validity of the model and provide new insights for HFSE.
- 1
- WWW 2008 workshop on social web search and mining1
SWSM2008 aimed at providing a forum for researchers from all over the world to share information on their latest investigations in SWN search, mining and its application particularly in academic research area.
- 1
- A survey of Web information system and applications1
The survey outlines the current activities and future trends concerning Web information systems and applications based on the papers accepted for publication in this issue of WISA.
- 1
- Research of Atomic and Anonymous Electronic Commerce Protocol1
Based on the concept of two-phase commitment, atomicity in electronic transaction with the Trusted Third Part as coordinator is realized, and Brands' fair signature model is developed, proposing a method to enable not only anonymity but also owner-trace and money-trace.
- 1
- List Data Extraction in Semi-structured Document1
A clustering based method making use of both the layout information and the semantic information of words for this extraction task is proposed, showing experimental results on plain-text annual reports from Shanghai Stock Exchange, in which 73.49% of the lists were extracted correctly.
- 1
- Research on Paging Strategy of Web Application1
This paper describes a computer simulation method which can be used to calculate the accessing probability of different page numbers and presents a paging framework which has already been practically used and which is fully configurable for different paging strategy.
- 1
- AISE-Bench: A Full-Cycle Curated Benchmark for Information Seeking on Academic Knowledge Graphs–
This work establishes a challenging new testbed for quantitatively evaluating and improving the stepwise correctness, grounded summarization, and traceable reasoning of multi-step API-using LLM agents, and develops a comprehensive evaluation protocol measuring answer quality, reference grounding, API-planning correctness, and execution success.
- –
- –
- –
- SurveyReview: A Reviewer-Aligned Benchmark for Survey Evaluators–
This work proposes SurveyReview, a reviewer-aligned, multi-dimensional benchmark and dataset for survey evaluation, and develops a strong baseline evaluator that substantially improves alignment with human reviewers, providing a competitive reference for future research.
- –
- Auxiliary Metrics Help Decoding Skill Neurons in the Wild–
This paper introduces a simple, lightweight, and broadly applicable method with a focus on isolating neurons that encode specific skills that can be correlated with auxiliary metrics -- such as external labels and the model's own confidence score -- thereby uncovering interpretable and task-specific behaviors without the need for manual token aggregation.
- –
- –
- –
- VidCoM: Fast Video Comprehension through Large Language Models with Multimodal Tools–
This paper introduces BiLL-VTG, a fast adaptive framework that leverages large language models (LLMs) to reasoning on videos based on essential lightweight visual tools and reveals the key to response specific instructions is the concentration on relevant video events.
- –
- –
- DSEL: A Domain-Specific Entity Linking System–
This paper proposes an unsupervised method to generate domain data from Wikipedia and provides a domain-specific neural collective entity linking model for each domain, and builds a system that can provide domain entity linking for users.
- –
- –
- Knowledge-Enhanced Bilingual Textual Representations for Cross-Lingual Semantic Textual Similarity–
It is shown that the proposed method outperforms cross-lingual representations methods using parallel data on cross-lingsual tasks, and achieves competitive results on mono-lingUAL tasks.
- Is a Common Phrase an Entity Mention or Not? Dual Representations for Domain-Specific Named Entity Recognition–
A novel BLSTM based NER model tailored for specific domains by learning dual representations for each word is presented, which learns not only general domain knowledge derived from an external large scale general corpus via a word embedding model, but also the specific domain knowledge by training a stacked deep neural network (SDNN) integrating the results of a low-cost pre-entity-linking process.
- –
- –
- –
- 基于非线性全局上下文的词嵌入–
A method for word representation that is based on nonlinear global context that can find a related word that is closer to the current meaning of the word, and it is better than the Word2Vec method in the word-sense disambiguation.
- –
- –
- –
- Enhance Social Context Understanding with Semantic Chunks–
This work proposes a method that extracts semantic chunks using these relations between importance words in formal texts to express social contexts and exploits semantic chunks by utilizing knowledge learned from semantically parsed corpora and knowledge base.
- –
- Term Committee Based Event Identification and Dependency Discovery–
The experimental results on two Linguistic Data Consortium datasets show that both the proposed methods for event identification and event dependency discovery have significant improvement over previous methods.
- Chinese statistical parser based on semantic dependencies–
A new tagging method is presented to build a Chinese semantic corpus that characterizes the sentence meaning as a linear sequence of dependency relationships which are the semantic or syntactic relationships between words in the sentence.
- –
- Design and implementation of data exchange bus based on user rule–
Experiment results show that rule-oriented ESB can perform data exchange perfectly and improve the reliability of data exchange.
- A Case Study of Multi-view News Exploration:Shanghai World Expo–
Through the NewsInsight system, people can know important events and entities reported during 2010 Shanghai World Expo from multi-view, and read the news of their interest from different views.
- –
- –
- –
- Translating OWL Specified Domain Knowledge to Aspect Oriented Model.–
A process for separating an urban waste mixture of paper and plastic film into its primary components comprising partially triturating the mixture to reduce the paper components to a small particle size ground condition.
- –
- Improved retrospective event detection algorithm–
Experimental results on news data from Xinhua News agency show that both the proposed approaches can effectively improve the performance of RED task,compare to the baseline method.
- The XML Schema Modeling for Radiation Environment Domain–
Based on the database management requirements in Nuclear Radiation Environment and the latest developments in database construction, XML Schema technology to the construction of the Data Model for Radiation Environment is applied and a real application for the Nuclear Radiation Database is established.
- SWARMS: A New Tool for Domain Exploration in Semantic Web–
SWARMS architecture, component technologies and its performance in FOAF domain will be described, because FOAF(Friend-Of-A-Friend) domain?s data size is largest among the three domains.
- Research and Implementation of Transformation Mechanism from 3D-model to 2D Vector Graphics–
This paper presents an improved algorithm of hidden component removal, for curve faces in 3D space, including a unified method of 3D modeling and a method of layer aggregation.
- Solution for J2EE System's Internationalization–
This solution has been applied to internationalize a J2EE based travel e-business system successfully and is proposed as a complete set of solution towards the internationalization of J2ee system.
- –
- Untagged Table Extraction in Semi-structured Documents–
An extraction algorithm is proposed by using its structural distribution features in documents to split the untagged table into rows and columns, and then inducts headers and merges cells.
- Knowledge Sharing by Grid Technology–
A framework for knowledge sharing by web service and grid technology is discussed and how other knowledge bases can be integrated into this framework and share their knowledge is mentioned.
- Research of Satisfying Atomic and Anonymous Electronic Commerce Protocol–
This paper proposes an approach to realize satisfying atomicity based on AFAP(Atomic and Fair Anonymous Protocol), and based on Brands’ fair signature model, this method satisfies both anonymity and owner-trace or money-trace.
- –
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.