Natural Language Processing and Chinese Computing: 12th National CCF Conference, NLPCC 2023, Foshan, China, October 12–15, 2023, Proceedings, Part III (Lecture Notes in Artificial Intelligence)

✍ Scribed by Fei Liu (editor), Nan Duan (editor), Qingting Xu (editor), Yu Hong (editor)

• Publisher: Springer
• Year: 2023
• Tongue: English
• Leaves: 438
• Category: Library

✦ Synopsis

This three-volume set constitutes the refereed proceedings of the 12th National CCF Conference on Natural Language Processing and Chinese Computing, NLPCC 2023, held in Foshan, China, during October 12–15, 2023.
The ____ regular papers included in these proceedings were carefully reviewed and selected from 478 submissions. They were organized in topical sections as follows: dialogue systems; fundamentals of NLP; information extraction and knowledge graph; machine learning for NLP; machine translation and multilinguality; multimodality and explainability; NLP applications and text mining; question answering; large language models; summarization and generation; student workshop; and evaluation workshop.

✦ Table of Contents

Preface
Organization
Contents – Part III
Poster: Summarization and Generation
Fantastic Gradients and Where to Find Them: Improving Multi-attribute Text Style Transfer by Quadratic Program
1 Introduction
2 Related Work
3 Methodology
3.1 Preliminary
3.2 Problem Analysis
3.3 Model Architecture
3.4 Multiple-Attributes Gradient Iterative Modification
4 Experiments
4.1 Dataset
4.2 Baselines
4.3 Evaluations Metrics
4.4 Main Results
4.5 Ablation Study
5 Conclusions
References
TiBERT: A Non-autoregressive Pre-trained Model for Text Editing
1 Introduction
2 Related Work
2.1 Text Editing Methods
2.2 Pre-trained Language Models
3 Method
3.1 Encoder
3.2 Locator
3.3 Editor
3.4 Pre-training
3.5 Fine-Tuning
4 Experiments
4.1 Settings
4.2 Data Conversion
4.3 Grammatical Error Correction (GEC)
4.4 Text Simplification (TS)
4.5 Chinese Spelling Check (CSC)
5 Analysis
6 Conclusion
References
Student Workshop: Information Retrieval and Text Mining
A Study on the Classification of Chinese Medicine Records Using BERT, Chest Impediment as an Example
1 Introduction
2 Data
2.1 Data Source
2.2 Data Labeling
3 Method
3.1 Pre-processing Stage
3.2 Pre-training Stage
3.3 Model Fine-tuning Stage
4 Results
4.1 Experimental Results of BERT Model
4.2 Experimental Results of Multi-model Comparison
5 Discuss
References
Student Workshop: Information Extraction and Knowledge Acquisition
Semantic Candidate Retrieval for Few-Shot Entity Linking
1 Introduction
2 Related Work
2.1 Candidate Generation
2.2 Entity Disambiguation
3 Semantic Candidate Retrieval
3.1 Candidate Generation
3.2 Entity Disambiguation
3.3 Post-processing
4 Experiments
4.1 Dataset
4.2 Baselines and Results
4.3 Analysis
5 Conclusion
References
ERNIE-AT-CEL: A Chinese Few-Shot Emerging Entity Linking Model Based on ERNIE and Adversarial Training
1 Introduction
2 Related Work
2.1 Pretrained Language Models
2.2 Candidate Entity Retrieval
3 Model Introduction
3.1 Task Definition
3.2 Model with ERNIE
3.3 Adversarial Training
4 Experiments
4.1 Preprocessing and Parameter Setting
4.2 Evaluation Metrics
4.3 Experimental Results
4.4 Analysis of Results
5 Conclusion
References
Evaluation Workshop: Chinese Grammatical Error Correction
HWCGEC:HW-TSC's 2023 Submission for the NLPCC2023's Chinese Grammatical Error Correction Task
1 Introduction
2 Related Work
3 Data Description
3.1 Schema Definition
3.2 Training Set
4 Methodology
4.1 Encoder-Decoder Model
4.2 Decoder Only Model
4.3 Distantly Supervised Data Augmentation
5 Experiment
5.1 Setup
5.2 Training Data
5.3 Development Data
5.4 Training Steps
5.5 Experimental Results
6 Conclusion
References
GrammarGPT: Exploring Open-Source LLMs for Native Chinese Grammatical Error Correction with Supervised Fine-Tuning
1 Introduction
2 Related Work
2.1 Grammatical Error Correction
2.2 Instruction Tuning for LLMs
3 Methods
3.1 Hybrid Dataset Construction
3.2 Error-invariant Data Augmentation
3.3 Instruction Tuning
4 Experiments
4.1 Datasets
4.2 Metrics
4.3 Hyper-parameters
4.4 Experimental Results
4.5 Ablation Study
5 Conclusion
References
Evaluation Workshop: Multi-perspective Scientific Machine Reading Comprehension
Enhanced CGSN System for Machine Reading Comprehension
1 Introduction
2 Related Work
3 The Proposed Approach
3.1 Task Definition
3.2 Model Structure
4 Experiments
4.1 Dataset and Metric
4.2 Experiment Settings
4.3 Baselines
4.4 Results and Analysis
5 Conclusion
References
Scientific Reading Comprehension with Sentences Selection and Ranking
1 Introduction
2 Related Work
3 Method
3.1 Retrieval Module
3.2 Re-Ranking Module
4 Experiments
4.1 Setuping
4.2 Tasks and Datasets
4.3 Evaluation Metrics
4.4 Main Results
5 Conclusion
References
Overview of NLPCC Shared Task 2: Multi-perspective Scientific Machine Reading Comprehension
1 Instruction
2 The Task
3 The Dataset
3.1 Data Format
4 Evaluation Metric
5 Participating Teams
6 Evaluation Results
7 Conclusion
References
Evaluation Workshop: Math Word Problem Solving
A Numeracy-Enhanced Decoding for Solving Math Word Problem
1 Introduction
2 Related Work
3 Methodology
3.1 Problem Statement
3.2 Basic Model Description
3.3 A Numeracy-Enhanced Token Embedding
3.4 A Numeracy-Enhanced Target Prediction
3.5 Training Objective
4 Experiments
4.1 Implementation Details
4.2 Datasets
4.3 Baselines
4.4 Evaluation and Results
4.5 Case Study
5 Conclusion
References
Solving Math Word Problem with Problem Type Classification
1 Introduction
2 Related Work
2.1 Ensemble Learning
2.2 Tree-Based MWP Solver
2.3 LLM Solver
3 Research Methodology
3.1 Problem Type Classifier
3.2 Bert2Tree Solver
3.3 LLM Solver
3.4 Post-processing
4 Experiments
4.1 Dataset Setting
4.2 Experimental Settings
4.3 Experimental Results
4.4 Case Study
5 Conclusion and Future Work
References
Consistent Solutions for Optimizing Search Space of Beam Search
1 Introduction
2 Related Work
3 Proposed System
3.1 Problem Statement
3.2 Consistent Solutions
3.3 Model Ensemble
4 Experiment
4.1 Dataset
4.2 Implementation Details
4.3 Experimental Results
4.4 Case Study
5 Conclusion
References
Evaluation Workshop: Conversational Aspect-Based Sentiment Quadruple Analysis
Improving Conversational Aspect-Based Sentiment Quadruple Analysis with Overall Modeling
1 Introduction
2 Related Work
2.1 Aspect-Based Sentiment Analysis
2.2 Dialogue Sentiment Analysis
2.3 Dialogue Aspect-Based Sentiment Analysis
3 Task Introduction
3.1 Task Definition
3.2 Evaluation Metric
4 Methodology
4.1 Overall Architecture
4.2 Text Representation Module
4.3 Multi-view Fusing Module
4.4 ConASQ Classification Module
5 Experiments
5.1 Dataset
5.2 Experimental Settings
5.3 Comparision Models
5.4 Main Results
5.5 Ablation Study
5.6 Online Results
6 Conclusion
References
Conversational Aspect-Based Sentiment Quadruple Analysis with Consecutive Multi-view Interaction
1 Introduction
2 Related Works
2.1 Aspect Sentiment Triplet Extraction
2.2 Emotion Recognization in Conversation
3 Methodology
3.1 Task Definition
3.2 Base Encoding
3.3 Consecutive Multi-view Interaction
3.4 Quadruple Decoding
3.5 Training Strategy
4 Experiment
4.1 Experimental Settings
4.2 Main Comparisons
4.3 Ablation Study
5 Conclusion
References
A Model Ensemble Approach for Conversational Quadruple Extraction
1 Introduction
2 Related Work
3 Methodology
3.1 Triplets Extraction Model
3.2 Models Selection
3.3 Ensemble Model
4 Experiments
4.1 Dataset and Evaluation Index
4.2 Results and Analysis
5 Conclusion
References
Enhancing Conversational Aspect-Based Sentiment Quadruple Analysis with Context Fusion Encoding Method
1 Introduction
2 Related Work
2.1 Aspect-Based Sentiment Quadruples Extraction
2.2 Conversational Aspect-Based Sentiment Quadruple Analysis
3 Methodology
3.1 Task Introduction
3.2 Context Fusion Encoding with Adversarial Training
3.3 Quadruple Decoder
3.4 Learning
4 Experiment
4.1 Datasets and Metrics
4.2 Experiment Setting
4.3 Baseline System
4.4 Results and Analysis
4.5 Effectiveness of Dialog Fusion
5 Conclusion
References
Evaluation Workshop: Chinese Medical Instructional Video Question Answering
A Unified Framework for Optimizing Video Corpus Retrieval and Temporal Answer Grounding: Fine-Grained Modality Alignment and Local-Global Optimization
1 Introduction
2 Related Work
2.1 Video Question Answering
2.2 Temporal Natural Language Localization in Video
3 Methodology
3.1 Problem Definition
3.2 Data Preprocess
3.3 Overall Architecture
4 Experimental Evaluation
4.1 Dataset and Evaluation Metrics
4.2 Experiment Settings
5 Conclusion
References
A Two-Stage Chinese Medical Video Retrieval Framework with LLM
1 Introduction
2 Related Work
3 The Proposed Approach
3.1 Task Definition
3.2 Method
4 Experiments
4.1 Dataset and Evaluation
4.2 Experimental Settings
4.3 Results and Discussions
5 Conclusion
References
Improving Cross-Modal Visual Answer Localization in Chinese Medical Instructional Video Using Language Prompts
1 Introduction
2 Related Work
2.1 Visual Answer Localization
2.2 Prompt Based Learning
3 Method
3.1 Task Formalization
3.2 Data Preprocess
3.3 Model Architecture
4 Experiments
4.1 Dataset and Metrics
4.2 Experiment Details
4.3 Experimental Results and Analysis
5 Conclusion
References
Overview of the NLPCC 2023 Shared Task: Chinese Medical Instructional Video Question Answering
1 Introduction
2 Task Introduction
2.1 Definition of Each Track
2.2 Evaluation Metrics
2.3 Dateset
3 Baseline Methods
3.1 Baseline Method for Track 1
3.2 Baseline Method for Track 2 and 3
4 Evaluation Results
5 Conclusion
References
Evaluation Workshop: Chinese Few-Shot and Zero-Shot Entity Linking
Improving Few-Shot and Zero-Shot Entity Linking with Coarse-to-Fine Lexicon-Based Retriever
1 Introduction
2 Related Work
3 Methodology
3.1 Coarse-to-Fine Lexicon-Based Retriever
3.2 BERT-Based Dual Encoder
3.3 Ensemble Method
4 Experiments
4.1 Data
4.2 Experimental Settings
4.3 Main Results
4.4 Performance of Retrieve Stage
4.5 Ablation Study
5 Conclusion
References
Overview of NLPCC 2023 Shared Task 6: Chinese Few-Shot and Zero-Shot Entity Linking
1 Introduction
2 Dataset
2.1 Knowledge Base
2.2 Few-Shot Evaluation Slice
2.3 Zero-Shot Evaluation Slice
3 Evaluation Results
3.1 Evaluation Systems
3.2 Submission Results
4 Conclusion
References
Evaluation Workshop: Chinese Essay Discourse Coherence Evaluation
Two-Stage Topic Sentence Extraction for Chinese Student Essays
1 Introduction
2 Related Work
3 Task Definition and Approach
3.1 Task Definition
3.2 Approach
4 Experiments
4.1 Evaluation Metrics
4.2 Data Analysis
4.3 Data Format Conversion
4.4 Experiment Settings
4.5 Results and Analysis
5 Conclusion
References
Multi-angle Prediction Based on Prompt Learning for Text Classification
1 Introduction
2 Related Work
2.1 BERT
2.2 Prompt Learning
2.3 Prompt Engineering
3 Task Definition
4 Method
4.1 Supervised Data Preprocessing Module
4.2 Prompt Characters Design and Template Building Module
4.3 Overview of Model Operation Flow
4.4 Training and Inference
5 Experiment
5.1 Dataset
5.2 Baseline
5.3 Implementation Details
5.4 Main Results
5.5 Analyze
6 Conclusion
References
Overview of the NLPCC 2023 Shared Task: Chinese Essay Discourse Coherence Evaluation
1 Introduction
2 Competition Tracks and Task Definitions
2.1 Coherence Evaluation (CE)
2.2 Text Topic Extraction (TTE)
2.3 Paragraph Logical Relation Recognition (PLRR)
2.4 Sentence Logical Relation Recognition (SLRR)
3 Dataset Description
3.1 Dataset Overview
3.2 Annotation Process
3.3 Dataset Distribution
4 Evaluation Metrics
5 Participated Systems
5.1 Track 1. Coherence Evaluation (CE)
5.2 Track 2. Text Topic Extraction (TTE)
5.3 Tracks 3 & 4: Paragraph-Level and Sentence-Level Logical Relation Recognition (PLRR & SLRR)
6 Results
7 Conclusions
References
Improving the Generalization Ability in Essay Coherence Evaluation Through Monotonic Constraints
1 Introduction
2 Related Works
3 Method
3.1 Local Discriminative Model
3.2 Punctuation Correction Model
3.3 Scorer
4 Experiments
4.1 Datasets
4.2 Experimental Settings
4.3 Results
5 Our Solution to the Remaining Tracks from NLPCC2023 Shared Task7
5.1 Text Topic Extraction (Track 2)
5.2 Paragraph Logical Relation Recognition (Track 3)
5.3 Sentence Logical Relation Recognition (Track 4)
6 Conclusion and Future Work
References
Task-Related Pretraining with Whole Word Masking for Chinese Coherence Evaluation
1 Introduction
2 Related Work
3 Method
4 Experiments
4.1 Dataset and Evaluation Metric
4.2 Implementation Details
4.3 Baselines
5 Results
6 Conclusion
References
Evaluation Workshop: Chinese Spelling Check
Towards Robust Chinese Spelling Check Systems: Multi-round Error Correction with Ensemble Enhancement
1 Introduction
2 Related Work
3 Our Approach
3.1 Data Preparation
3.2 Multi-round Error Correction
3.3 Two Strategies of Ensemble Enhancement
3.4 Sentence Calibration
4 Experiments
4.1 Datasets
4.2 Comparable Methods
4.3 Implementation Details
4.4 Comparison with Other Systems
4.5 Comparison with Variants
4.6 Case Study
5 Conclusions
References
Overview of the NLPCC 2023 Shared Task: Chinese Spelling Check
1 Introduction
2 Related Work
3 Dataset
3.1 Dataset Provided
3.2 Analysis of SIGHAN Datasets
4 Evaluation Metrics
4.1 Detection Level
4.2 Correction Level
4.3 False Positive Rate
5 Evaluation
5.1 Results
5.2 Result Analysis
5.3 Representative Systems
6 Conclusion
References
Evaluation Workshop: User Feedback Prediction and Response Generation
User Preference Prediction for Online Dialogue Systems Based on Pre-trained Large Model
1 Introduction
2 Related Work
3 Datasets
4 Model Implementation
4.1 Pre-training Large Model
4.2 Data Pre-processing
4.3 Model Selection
4.4 Data Analysis and Data Enhancement
4.5 Semi-supervised Learning with Pseudo-labeling
4.6 Model Fusion
5 Experimental Results
6 Summary
References
Auto-scaling Distribution Fitting Network for User Feedback Prediction
1 Introduction
2 Related Work
3 Main Methods
3.1 Smoothing Coefficient
3.2 Automatic Temperature Scaling Layer
4 Experiments
4.1 Data Analysis and Evaluation Metric
4.2 Experimental Settings
4.3 Main Results
4.4 Online Results
5 Conclusions
References
Adversarial Training and Model Ensemble for User Feedback Prediciton in Conversation System
1 Introduction
2 Related Work
3 Methodology
3.1 Task Definition
3.2 Model Architecture
3.3 Adversarial Training
3.4 Data Augmentation
3.5 Blending for Global Ensemble
4 Experiments
4.1 Dataset and Evaluation
4.2 Results and Analysis
5 Conclusion and Future Work
References
Generating Better Responses from User Feedback via Reinforcement Learning and Commonsense Inference
1 Introduction
2 Related Work
2.1 Dialogue Generation
2.2 Dialogue Based on User Feedback
2.3 Commonsense Inference in Dialogue
3 Methodology
3.1 Task Definition
3.2 Commonsense Inference
3.3 RLHF
3.4 Contrastive Search
4 Experiments
4.1 Dataset Description
4.2 Experimental Settings
4.3 Evaluation Metrics
4.4 Baselines
4.5 Automatic Evaluation
4.6 Ablation Study
4.7 Online Evaluation
5 Conclusions
References
Overview of the NLPCC 2023 Shared Task 9: User Feedback Prediction and Response Generation
1 Introduction
2 Related Work
3 Task Description
3.1 Track 1: Prediction of Likes and Dislikes
3.2 Track 2: Conversation Generation Based on Likes and Dislikes
4 Dataset Description
5 Results
5.1 Evaluation Metrics
5.2 Participants
5.3 Main Results
6 Conclusion
References
Evaluation Workshop: Learn to Watch TV: Multimodal Dialogue Understanding and Response Prediction
Multimodal Dialogue Understanding via Holistic Modeling and Sequence Labeling
1 Introduction
2 Related Work
2.1 Multimodal Dialogue Scene Identification
2.2 Multimodal Dialogue Session Identification
3 Task Introduction
3.1 Task Definition
3.2 Evaluation Metric
4 Methodology
4.1 Overall Architecture
4.2 Text Representation Module
4.3 Image Representation Module
4.4 Multimodal Fusing and Classification Module
5 Experiments
5.1 Dataset
5.2 Experimental Settings
5.3 Comparision Models
5.4 Main Results
5.5 Ablation Study
5.6 Online Results
6 Conclusion
References
Overview of the NLPCC 2023 Shared Task 10: Learn to Watch TV: Multimodal Dialogue Understanding and Response Generation
1 Introduction
2 Task Description
2.1 Dialogue Scene Identification
2.2 Dialogue Session Identification
2.3 Dialogue Response Retrieval
2.4 Dialogue Response Generation
3 Dataset Description
4 Results
5 Conclusion
References
Correction to: A Numeracy-Enhanced Decoding for Solving Math Word Problem
Correction to: Chapter “A Numeracy-Enhanced Decoding for Solving Math Word Problem” in: F. Liu et al. (Eds.): Natural Language Processing and Chinese Computing, LNAI 14304, https://doi.org/10.1007/978-3-031-44699-3_11
Author Index