A5025631695- Topic Modeling
- Recommender Systems and Techniques
- Natural Language Processing Techniques
- Web Data Mining and Analysis
- Advanced Graph Neural Networks
- Information Retrieval and Search Behavior
- Multimodal Machine Learning Applications
- Data Management and Algorithms
- Advanced Image and Video Retrieval Techniques
- Domain Adaptation and Few-Shot Learning
- Text and Document Classification Technologies
- Image Retrieval and Classification Techniques
- Advanced Database Systems and Queries
- Advanced Bandit Algorithms Research
- Speech and dialogue systems
- Semantic Web and Ontologies
- Complex Network Analysis Techniques
- Algorithms and Data Compression
- Advanced Text Analysis Techniques
- Expert finding and Q&A systems
- Caching and Content Delivery
- Data Quality and Management
- Machine Learning in Healthcare
- Video Analysis and Summarization
- Speech Recognition and Synthesis
Renmin University of China
2016-2025
Beijing Institute of Big Data Research
2015-2023
IT University of Copenhagen
2023
Tokyo Institute of Technology
2023
Administration for Community Living
2023
American Jewish Committee
2023
Baidu (China)
2023
Data Management (Italy)
2022
Beijing Academy of Artificial Intelligence
2021-2022
Beijing University of Posts and Telecommunications
2022
Language is essentially a complex, intricate system of human expressions governed by grammatical rules. It poses significant challenge to develop capable AI algorithms for comprehending and grasping language. As major approach, language modeling has been widely studied understanding generation in the past two decades, evolving from statistical models neural models. Recently, pre-trained (PLMs) have proposed pre-training Transformer over large-scale corpora, showing strong capabilities...
Large-scale pre-trained models (PTMs) such as BERT and GPT have recently achieved great success become a milestone in the field of artificial intelligence (AI). Owing to sophisticated pre-training objectives huge model parameters, large-scale PTMs can effectively capture knowledge from massive labeled unlabeled data. By storing into parameters fine-tuning on specific tasks, rich implicitly encoded benefit variety downstream which has been extensively demonstrated via experimental...
Recently, significant progress has been made in sequential recommendation with deep learning. Existing neural models usually rely on the item prediction loss to learn model parameters or data representations. However, trained this is prone suffer from sparsity problem. Since it overemphasizes final performance, association fusion between context and sequence not well captured utilized for recommendation. To tackle problem, we propose S^3-Rec, which stands Self-Supervised learning Sequential...
Although personalized search has been proposed for many years and personalization strategies have investigated, it is still unclear whether consistently effective on different queries users, under contexts. In this paper, we study problem get some preliminary conclusions. We present a large-scale evaluation framework based query logs, then evaluate five (including two click-based three profile-based ones) using 12-day MSN logs. By analyzing the results, reveal that significant improvement...
Query expansion has long been suggested as an effective way to resolve the short query and word mismatching problems. A number of methods have proposed in traditional information retrieval. However, these previous do not take into account specific characteristics web searching; particular, availability large amount user interaction recorded logs. In this study, we propose a new method for based on The central idea is extract probabilistic correlations between terms document by analyzing...
Article Share on Clustering user queries of a search engine Authors: Ji-Rong Wen Microsoft Research, China, 5F, Beijing Sigma Center, No.49, Zhichun Road Haidian District, Beijing, P.R.China P.R.ChinaView Profile , Jian-Yun Nie Dept. Informatique et Recherche, opérationnelle, University Montreal, CP 6128, succursale Centre-ville, H3C 3J7 Canada CanadaView Hong-Jiang Zhang Authors Info & Claims WWW '01: Proceedings the 10th international conference World Wide WebMay 2001 Pages...
With the revival of neural networks, many studies try to adapt powerful sequential models, ıe Recurrent Neural Networks (RNN), recommendation. RNN-based networks encode historical interaction records into a hidden state vector. Although vector is able dependency, it still has limited representation power in capturing complicated user preference. It difficult capture fine-grained preference from sequence. Furthermore, latent usually hard understand and explain. To address these issues, this...
Query clustering is a process used to discover frequently asked questions or most popular topics on search engine. This crucial for engines based question-answering. Because of the short lengths queries, approaches keywords are not suitable query clustering. paper describes new method that makes use user logs which allow us identify documents users have selected query. The similarity between two queries may be deduced from common them. Our experiments show combination both and better than...
VQA models may tend to rely on language bias as a shortcut and thus fail sufficiently learn the multi-modal knowledge from both vision language. Recent debiasing methods proposed exclude prior during inference. However, they disentangle "good" context "bad" whole. In this paper, we investigate how mitigate in VQA. Motivated by causal effects, novel counterfactual inference framework, which enables us capture direct effect of questions answers reduce subtracting total effect. Experiments...
In recent years, there are a large number of recommendation algorithms proposed in the literature, from traditional collaborative filtering to deep learning algorithms. However, concerns about how standardize open source implementation continually increase research community. light this challenge, we propose unified, comprehensive and efficient recommender system library called RecBole (pronounced as [rEk'[email protected]]), which provides unified framework develop reproduce for purpose....
Conversational recommender systems (CRS) aim to recommend high-quality items users through interactive conversations. Although several efforts have been made for CRS, two major issues still remain be solved. First, the conversation data itself lacks of sufficient contextual information accurately understanding users' preference. Second, there is a semantic gap between natural language expression and item-level user
In recent years, the boundaries between e-commerce and social networking have become increasingly blurred. Many Web sites support mechanism of login where users can sign on using their network identities such as Facebook or Twitter accounts. Users also post newly purchased products microblogs with links to product pages. this paper, we propose a novel solution for cross-site cold-start recommendation, which aims recommend from at in "cold-start" situations, problem has rarely been explored...
Abstract Autonomous agents have long been a research focus in academic and industry communities. Previous often focuses on training with limited knowledge within isolated environments, which diverges significantly from human learning processes, makes the hard to achieve human-like decisions. Recently, through acquisition of vast amounts Web knowledge, large language models (LLMs) shown potential human-level intelligence, leading surge LLM-based autonomous agents. In this paper, we present...
Multi-hop Knowledge Base Question Answering (KBQA) aims to find the answer entities that are multiple hops away in Knowl- edge (KB) from question. A major challenge is lack of supervision signals at intermediate steps. Therefore, multi-hop KBQA algorithms can only receive feedback final answer, which makes learning unstable or ineffective. To address this challenge, we propose a novel teacher-student approach for task. In our approach, stu- dent network correct query, while teacher tries...
Abstract The fundamental goal of artificial intelligence (AI) is to mimic the core cognitive activities human. Despite tremendous success in AI research, most existing methods have only single-cognitive ability. To overcome this limitation and take a solid step towards general (AGI), we develop foundation model pre-trained with huge multimodal data, which can be quickly adapted for various downstream tasks. achieve goal, propose pre-train our by self-supervised learning weak semantic...
Inspired by the superior language abilities of large models (LLM), vision-language (LVLM) have been recently proposed integrating powerful LLMs for improving performance on complex multimodal tasks. Despite promising progress LVLMs, we find that they suffer from object hallucinations, i.e., tend to generate objects inconsistent with target images in descriptions. To investigate it, this work presents first systematic study hallucination LVLMs. We conduct evaluation experiments several...
In various natural language processing tasks, passage retrieval and re-ranking are two key procedures in finding ranking relevant information. Since both the contribute to final performance, it is important jointly optimize them order achieve mutual improvement. this paper, we propose a novel joint training approach for dense reranking. A major contribution that introduce dynamic listwise distillation, where design unified retriever re-ranker. During re-ranker can be adaptively improved...
Knowledge base question answering (KBQA) aims to answer a over knowledge (KB). Recently, large number of studies focus on semantically or syntactically complicated questions. In this paper, we elaborately summarize the typical challenges and solutions for complex KBQA. We begin with introducing background about KBQA task. Next, present two mainstream categories methods KBQA, namely semantic parsing-based (SP-based) information retrieval-based (IR-based) methods. then review advanced...
In order to develop effective sequential recommenders, a series of sequence representation learning (SRL) methods are proposed model historical user behaviors. Most existing SRL rely on explicit item IDs for developing the models better capture preference. Though some extent, these difficult be transferred new recommendation scenarios, due limitation by explicitly modeling IDs. To tackle this issue, we present novel universal approach, named UniSRec. The approach utilizes associated...
Sequential recommendation aims at predicting users' preferences based on their historical behaviors. However, this strategy may not perform well in practice due to the sparsity of real-world data. In paper, we propose a novel counterfactual data augmentation framework mitigate impact imperfect training and empower sequential models. Our is composed sampler model an anchor model. The generate new user behavior sequences observed ones, while leveraged provide final list, which trained both...
Large language models (LLMs), such as ChatGPT, are prone to generate hallucinations, i.e., content that conflicts with the source or cannot be verified by factual knowledge. To understand what types of and which extent LLMs apt hallucinate, we introduce Hallucination Evaluation for Language Models (HaluEval) benchmark, a large collection generated human-annotated hallucinated samples evaluating performance in recognizing hallucination. these samples, propose ChatGPT-based two-step framework,...
Conversational recommender systems (CRS) aim to proactively elicit user preference and recommend high-quality items through natural language conversations. Typically, a CRS consists of recommendation module predict preferred for users conversation generate appropriate responses. To develop an effective CRS, it is essential seamlessly integrate the two modules. Existing works either design semantic alignment strategies, or share knowledge resources representations between However, these...
Recently, contrastive learning has been shown to be effective in improving pre-trained language models (PLM) derive high-quality sentence representations. It aims pull close positive examples enhance the alignment while push apart irrelevant negatives for uniformity of whole representation space.However, previous works mostly adopt in-batch or sample from training data at random. Such a way may cause sampling bias that improper (false and anisotropy representations) are used learn...
Autonomous agents have long been a prominent research focus in both academic and industry communities. Previous this field often focuses on training with limited knowledge within isolated environments, which diverges significantly from human learning processes, thus makes the hard to achieve human-like decisions. Recently, through acquisition of vast amounts web knowledge, large language models (LLMs) demonstrated remarkable potential achieving human-level intelligence. This has sparked an...