A5102761202- Topic Modeling
- Advanced Text Analysis Techniques
- Natural Language Processing Techniques
- Sentiment Analysis and Opinion Mining
- Mental Health via Writing
- Artificial Intelligence in Law
- Emotion and Mood Recognition
- Text and Document Classification Technologies
- Advanced Graph Neural Networks
- Data Mining Algorithms and Applications
- Computational and Text Analysis Methods
- Privacy, Security, and Data Protection
- Ethics and Social Impacts of AI
- Flow Measurement and Analysis
- Dispute Resolution and Class Actions
- Traffic Prediction and Management Techniques
- Recommender Systems and Techniques
- Interpreting and Communication in Healthcare
- Speech Recognition and Synthesis
- Machine Learning in Healthcare
- FinTech, Crowdfunding, Digital Finance
- Personality Traits and Psychology
- Law, Economics, and Judicial Systems
- Blockchain Technology Applications and Security
- Humor Studies and Applications
Hong Kong Polytechnic University
2021-2024
Huawei Technologies (China)
2024
University of Science and Technology Beijing
2020-2022
Abstract Distributed acoustic sensing (DAS) technology is a fiber-optic based distributed technology. It achieves real-time monitoring of signals by detecting weak disturbances along the fiber. has advantages such as long measurement distance, high spatial resolution and large dynamic range. Artificial intelligence (AI) great application potential in DAS technology, including data augmentation, preprocessing classification recognition events. By introducing AI algorithms, system can process...
To provide consistent emotional interaction with users, dialog systems should be capable to automatically select appropriate emotions for responses like humans.However, most existing works focus on rendering specified in or empathetically respond the emotion of yet individual difference expression is overlooked.This may lead inconsistent expressions and disinterest users.To tackle this issue, we propose equip system personality enable it by simulating transition humans conversation.In...
Writing a survey paper on one research topic usually needs to cover the salient content from numerous related papers, which can be modeled as multi-document summarization (MDS) task. Existing MDS datasets focus producing structureless summary covering few input documents. Meanwhile, previous structured generation works summarizing single document into multi-section summary. These existing and methods cannot meet requirements of academic papers To deal with scarcity available data, we propose...
Abstractive summarization aims to generate a concise summary covering the input document's salient information. Within report document, information can be scattered in textual and non-textual content. However, existing document datasets methods usually focus on text filter out Missing tabular data limit produced summaries' informativeness, especially when summaries require quantitative descriptions of critical metrics tables. Existing cannot meet requirements summarizing long dozens tables...
Abstractive multi-document summarization aims to generate a comprehensive summary covering salient content from multiple input documents.Compared with previous RNNbased models, the Transformer-based models employ self-attention mechanism capture dependencies in documents and can better summaries.Existing works have not considered key phrases determining attention weights of self-attention.Consequently, some tokens within only receive small weights.It affect completely encoding that convey...
Automatic document summarization aims to produce a concise summary covering the input document’s salient information. Within report document, information can be scattered in textual and non-textual content. However, existing datasets methods usually focus on text filter out Missing tabular data limit produced summaries’ informativeness, especially when summaries require quantitative descriptions of critical metrics tables. Existing cannot meet requirements summarizing long multiple tables...
In the research of P2P lending data, study borrower characteristics is great value for establishment target customers and risk management. Because high dimensionality, mixed attributes, different importance, generation time information, data often leads to mining results unable reflect important that affect approval loan amount. this article, we are first propose attributes partition considering business process classify variables into types. Furthermore, a multiangle method by discover...
Common law courts need to refer similar precedents' judgments inform their current decisions. Generating high-quality summaries of court judgment documents can facilitate legal practitioners efficiently review previous cases and assist the general public in accessing how operate is applied. Previous summarization research focuses on civil or a particular jurisdiction's judgments. However, judges from all common jurisdictions. Current datasets are insufficient satisfy demands summarizing...
Generating appropriate emotions for responses is essential dialogue systems to provide human-like interaction in various application scenarios. Most previous tried achieve this goal by learning empathetic manners from anonymous conversational data. However, emotional generated those methods may be inconsistent, which will decrease user engagement and service quality. Psychological findings suggest that the expressions of humans are rooted personality traits. Therefore, we propose a new task,...
Personality Recognition in Conversation (PRC) aims to identify the personality traits of speakers through textual dialogue content. It is essential for providing personalized services various applications Human-Computer Interaction (HCI), such as AI-based mental therapy and companion robots elderly. Most recent studies analyze dialog content classification yet overlook two major concerns that hinder their performance. First, crucial implicit factors contained conversation, emotions reflect...
Generating appropriate emotions for responses is essential dialog systems to provide human-like interaction in various application scenarios. Most previous tried achieve this goal by learning empathetic manners from anonymous conversational data. However, emotional generated those methods may be inconsistent, which will decrease user engagement and service quality. Psychological findings suggest that the expressions of humans are rooted personality traits. Therefore, we propose a new task,...
Recent studies successfully learned static graph embeddings that are structurally fair by preventing the effectiveness disparity of high- and low-degree vertex groups in downstream mining tasks. However, achieving structure fairness dynamic embedding remains an open problem. Neglecting degree changes graphs will significantly impair without notably improving fairness. This is because performance high-degree low-to-high-degree vertices drop close to generally poorer most slightly changed...
Personality Recognition in Conversation (PRC) aims to identify the personality traits of speakers through textual dialogue content. It is essential for providing personalized services various applications Human-Computer Interaction (HCI), such as AI-based mental therapy and companion robots elderly. Most recent studies analyze dialog content classification yet overlook two major concerns that hinder their performance. First, crucial implicit factors contained conversation, emotions reflect...
This work is motivated by two key trends. On one hand, large language models (LLMs) have shown remarkable versatility in various generative tasks such as writing, drawing, and question answering, significantly reducing the time required for many routine tasks. other researchers, whose not only time-consuming but also highly expertise-demanding, face increasing challenges they to spend more reading, reviewing papers. raises question: how can LLMs potentially assist researchers alleviating...
Recent studies successfully learned static graph embeddings that are structurally fair by preventing the effectiveness disparity of high- and low-degree vertex groups in downstream mining tasks. However, achieving structure fairness dynamic embedding remains an open problem. Neglecting degree changes graphs will significantly impair without notably improving fairness. This is because performance high-degree low-to-high-degree vertices drop close to generally poorer most slightly changed...
Aspect-based meeting transcript summarization aims to produce multiple summaries, each focusing on one aspect of content in a transcript. It is challenging as sentences related different aspects can mingle together, and those relevant specific be scattered throughout the long meeting. The traditional methods summary mixing information all aspects, which cannot deal with above challenges aspect-based summarization. In this paper, we propose two-stage method for To select input train sentence...
Task-Oriented Parsing (TOP) enables conversational assistants to interpret user commands expressed in natural language, transforming them into structured outputs that combine elements of both language and intent/slot tags. Recently, Large Language Models (LLMs) have achieved impressive performance synthesizing computer programs based on a prompt, mitigating the gap between programs. Our paper focuses harnessing capabilities LLMs for semantic parsing tasks, addressing following three key...
Recommending loan products to applicants would benefit many financial businesses and individuals. Nevertheless, suffer from the cold start problem; i.e., there are no available historical data for training recommendation model. Considering delayed feedback complex semantic properties of loans, methods general cannot be directly used. Moreover, existing ignore default risk, which should evaluated along with approval rate. To solve these challenges, we propose CSRLoan recommendation. employs...