A5002982984- Topic Modeling
- Natural Language Processing Techniques
- Advanced Text Analysis Techniques
- Text Readability and Simplification
- Sentiment Analysis and Opinion Mining
- Educational and Psychological Assessments
- Advanced Graph Neural Networks
- Counseling, Therapy, and Family Dynamics
- Machine Learning in Healthcare
- Child and Adolescent Psychosocial and Emotional Development
- Cognitive Science and Mapping
- Multimodal Machine Learning Applications
- Text and Document Classification Technologies
- Education and Critical Thinking Development
- Law, AI, and Intellectual Property
- Advanced Image and Video Retrieval Techniques
- Psychological Treatments and Assessments
- Semantic Web and Ontologies
- Mental Health Research Topics
- Artificial Intelligence in Healthcare and Education
- Handwritten Text Recognition Techniques
- Image Retrieval and Classification Techniques
Shenzhen Institutes of Advanced Technology
2023-2024
Chinese Academy of Sciences
2023-2024
Singapore Management University
2020-2022
Allen Institute for Artificial Intelligence
2022
East Stroudsburg University
2022
Ohio University
2022
University of North Carolina at Charlotte
2022
Carnegie Mellon University
2022
Salesforce (United States)
2022
In this paper, we revisit math word problems (MWPs) from the cross-lingual and multilingual perspective. We construct our MWP solvers over pretrained language models using sequence-to-sequence model with copy mechanism. compare how perform in scenarios. To facilitate comparison of performance, first adapt large-scale English dataset MathQA as a counterpart Chinese Math23K. Then extend several datasets to bilingual through machine translation plus human annotation. Our experiments show that...
Traditional in-person psychological counseling remains primarily niche, often chosen by individuals with issues, while online automated offers a potential solution for those hesitant to seek help due feelings of shame. Cognitive Behavioral Therapy (CBT) is an essential and widely used approach in counseling. The advent large language models (LLMs) agent technology enables automatic CBT diagnosis treatment. However, current LLM-based systems use agents fixed structure, limiting their...
Large language models (LLMs) have demonstrated impressive performance in various natural processing (NLP) tasks. However, there is limited understanding of how well LLMs perform specific domains (e.g, the intellectual property (IP) domain). In this paper, we contribute a new benchmark, first Multilingual-oriented quiZ on Intellectual Property (MoZIP), for evaluation IP domain. The MoZIP benchmark includes three challenging tasks: multiple-choice quiz (IPQuiz), question answering (IPQA), and...
Using large language models (LLMs) to assist psychological counseling is a significant but challenging task at present. Attempts have been made on improving empathetic conversations or acting as effective assistants in the treatment with LLMs. However, existing datasets lack consulting knowledge, resulting LLMs lacking professional competence. Moreover, how automatically evaluate multi-turn dialogues within process remains an understudied area. To bridge gap, we propose CPsyCoun,...
People perceive the world with multiple senses (e.g., through hearing sounds, reading words and seeing objects). However, most existing AI systems only process an individual modality. This paper presents approach that excels at handling modalities of information a single model. In our "{SkillNet}" model, different parts parameters are specialized for processing modalities. Unlike traditional dense models always activate all model parameters, sparsely activates whose skills relevant to task....
Understanding idioms is important in NLP.In this paper, we study to what extent a pretrained BERT model able encode the meaning of potentially idiomatic expression (PIE) certain context.We make use few existing datasets and perform two probing tasks: PIE usage classification idiom paraphrase identification.Our experiment results suggest that indeed separate literal usages with high accuracy.It also some extent.
In Chinese, Chengyu are fixed phrases consisting of four characters. As a type idioms, their meanings usually cannot be derived from component this article, we study the task recommending given textual context. Observing some limitations with existing work, propose two-stage model, where during first stage re-train Chinese BERT model by masking out large corpus wide coverage Chengyu. During second stage, fine-tune re-trained, Chengyu-oriented on specific recommendation dataset. We evaluate...
Chinese idioms are special fixed phrases usually derived from ancient stories, whose meanings oftentimes highly idiomatic and non-compositional. The idiom prediction task is to select the correct a set of candidate given context with blank. We propose BERT-based dual embedding model encode contextual words as well learn embeddings idioms. Specifically, we first match each hidden representation corresponding blank in context. then representations all tokens thorough pooling. further use two...
Minghuan Tan, Yong Dai, Duyu Tang, Zhangyin Feng, Guoping Huang, Jing Jiang, Jiwei Li, Shuming Shi. Proceedings of the 60th Annual Meeting Association for Computational Linguistics (Volume 1: Long Papers). 2022.
In this paper, we introduce a novel psychological benchmark, CPsyExam, constructed from questions sourced Chinese language examinations. CPsyExam is designed to prioritize knowledge and case analysis separately, recognizing the significance of applying real-world scenarios. From pool 22k questions, utilize 4k create benchmark that offers balanced coverage subjects incorporates diverse range techniques.Furthermore, evaluate existing large models~(LLMs), spanning open-sourced API-based models....
Numeral systems and units of measurement are two conjoined topics in activities human beings have mutual effects with the languages expressing them. Currently, evaluation Large Language Models (LLMs) often involves mathematical reasoning, yet little attention is given to how minor changes numbers or can drastically alter complexity problems performance LLMs. In this paper, we scrutinize existing LLMs on processing numerals by constructing datasets perturbations. We first anatomize reasoning...
The rapid progress in Large Language Models (LLMs) has prompted the creation of numerous benchmarks to evaluate their capabilities.This study focuses on Comprehensive Medical Benchmark Chinese (CMB), showcasing how dataset diversity and distribution supervised fine-tuning (SFT) may enhance LLM performance.Remarkably, We successfully trained a smaller base model achieve scores comparable larger models, indicating that diverse well-distributed can optimize performance regardless size.This...
Empathetic response generation is designed to comprehend the emotions of others and select most appropriate strategies assist them in resolving emotional challenges. Empathy can be categorized into cognitive empathy affective empathy. The former pertains ability understand discern issues situations others, while latter involves capacity provide comfort. To enhance one's empathetic abilities, it essential develop both these aspects. Therefore, we an innovative framework that combines...
Idioms represent a ubiquitous vehicle for conveying sentiments in the realm of everyday discourse, rendering nuanced analysis idiom sentiment crucial comprehensive understanding emotional expression within real-world texts. Nevertheless, existing corpora dedicated to considerably limit research text analysis. In this paper, we propose an innovative approach automatically expand lexicon idioms, leveraging capabilities large language models through application Chain-of-Thought prompting. To...
This work presents the task of text polishing, which generates a sentence that is more graceful than input while retaining its semantic meaning. Text polishing has great value in real usage and an important component modern writing assistance systems. However, still not well studied literature. Further research this direction requires formal definitions, benchmark datasets, powerful baseline models. In work, we formulate as context-dependent generation problem conduct case study on with...
In this paper, we investigate how to improve tagging-based Grammatical Error Correction models.We address two issues of current approaches, label imbalance issue, and tagging entanglement issue.Then propose down-weight the loss well-classified labels using Focal Loss decouple error detection layer from through an extra self-attention-based matching module.Experiments over three latest Chinese datasets show that our proposed methods are effective.We further analyze choices hyper-parameters...
In this paper, we revisit math word problems~(MWPs) from the cross-lingual and multilingual perspective. We construct our MWP solvers over pretrained language models using sequence-to-sequence model with copy mechanism. compare how perform in scenarios. To facilitate comparison of performance, first adapt large-scale English dataset MathQA as a counterpart Chinese Math23K. Then extend several datasets to bilingual through machine translation plus human annotation. Our experiments show that...
We study the task of learning and evaluating Chinese idiom embeddings.We first construct a new evaluation dataset that contains synonyms antonyms.Observing existing word embedding methods may not be suitable for embeddings, we further present BERT-based method directly learns vectors individual idioms.We empirically compare representative our method.We find substantially outperforms on have constructed.
Chinese idioms are special fixed phrases usually derived from ancient stories, whose meanings oftentimes highly idiomatic and non-compositional. The idiom prediction task is to select the correct a set of candidate given context with blank. We propose BERT-based dual embedding model encode contextual words as well learn embeddings idioms. Specifically, we first match each hidden representation corresponding blank in context. then representations all tokens thorough pooling. further use two...
This paper describes an approach to detect idiomaticity only from the contextualized representation of a MWE over multilingual pretrained language models.Our experiments find that larger models are usually more effective in detection. However, using higher layer model may not guarantee better performance.In scenarios, convergence different languages consistent and rich-resource have big advantages other languages.
While GPT has become the de-facto method for text generation tasks, its application to pinyin input remains unexplored. In this work, we make first exploration leverage Chinese method. We find that a frozen achieves state-of-the-art performance on perfect pinyin. However, drops dramatically when includes abbreviated A reason is an can be mapped many pinyin, which links even larger number of characters. mitigate issue with two strategies, including enriching context and optimizing training...