Machine Translation: 19th China Conference, CCMT 2023, Jinan, China, October 19–21, 2023, Proceedings (Communications in Computer and Information Science)

This book constitutes the refereed proceedings of the 19th China Conference on Machine Translation, CCMT 2023, held in Jinan, China, during October 19–21, 2023. The 8 full papers and 3 short papers included in this book were carefully reviewed and selected from 71 submissions. They focus on machine translation; improvement of translation models and systems; translation quality estimation; document-level machine translation; low-resource machine translation. Preface Organization Contents Transn's Submission for CCMT 2023 Quality Estimation Task 1 Introduction 2 Related Work 3 Feature-Enhanced Estimator for Sentence-Level QE 3.1 Model Architecture 3.2 Pretraining Corpus Generation 3.3 Model Ensemble 4 Experiments 4.1 Datasets 4.2 Training and Evaluation 4.3 Results and Analysis 4.4 Model Ensemble 5 Conclusion References HW-TSC's Neural Machine Translation System for CCMT 2023 1 Introduction 2 Dataset 2.1 Data Size 2.2 Data Pre-processing 3 System Overview 3.1 Bilingual System 3.2 Low-Resource System 3.3 Multilingual System 3.4 Zero-Referencing System 4 Method 4.1 Regularized Dropout 4.2 Bidirectional Training 4.3 Data Diversification 4.4 Forward Translation 4.5 Back-Translation 4.6 Alternated Training 4.7 Curriculum Learning 4.8 Transductive Ensemble Learning 5 Experiments 5.1 Bilingual System Evaluation Results 5.2 Low-Resource System Evaluation Results 5.3 Multilingual System Evaluation Results 5.4 Zero-Referencing System Evaluation Results 6 Conclusion References CCMT2023 Tibetan-Chinese Machine Translation Evaluation Technical Report 1 Introduction 2 Data Processing 2.1 Data 2.2 Data Preprocessing 3 Model 3.1 Model Select 3.2 Model Ensemble 3.3 Iterative Fine-Tuning 4 Experiment 4.1 Experimental Environment 4.2 Experimental Setup 5 Results and Analysis 6 Summary References Korean-Chinese Machine Translation Method Based on Independent Language Features 1 Introduction 2 Related Work 2.1 Korean-to-Chinese Machine Translation 2.2 Multilingual Unsupervised and Supervised Embeddings 3 Method 3.1 Independent Language Feature Extraction Model 3.2 Translation Model 4 Experiment 4.1 Datasets 4.2 Settings 4.3 Main Results 5 Analysis 5.1 Ablation Experiments 5.2 Case Study 6 Conclusion References NJUNLP's Submission for CCMT 2023 Quality Estimation Task 1 Introduction 2 Methods 2.1 Unsupervised Methods 2.2 Supervised Methods 3 Experiments 3.1 Dataset 3.2 Settings 3.3 Single Model Results 3.4 Ensemble 4 Conclusion References HIT-MI&T Lab's Submission to CCMT 2023 Automatic Post-editing Task 1 Introduction 2 Architecture 3 Data Augmentation 3.1 Synthetic Data Generation 3.2 ChatGPT-Based Data Augmentation 4 Experiments 4.1 Set-Up 4.2 Results of Different Architectures 4.3 Results of Data Augmentation 4.4 Results of Multi-model Ensemble 5 Conclusion References A k-Nearest Neighbor Approach for Domain-Specific Translation Quality Estimation 1 Introduction 2 Proposed Method 2.1 Overall Architecture 2.2 XLM-R Encoder 2.3 Classifier 2.4 k-Nearest Neighbor 2.5 Loss Function 3 Experiments 3.1 Datasets 3.2 Settings 3.3 Results and Analysis 3.4 Results After Integrating Domain Transfer Method 3.5 Effects of the KNN Parameters 4 Conclusion References WSA: A Unified Framework for Word and Sentence Autocompletion in Interactive Machine Translation 1 Introduction 2 Related Work 3 Proposed Method 3.1 Task Definition 3.2 The Joint Model 3.3 The Use of Initial Translation 3.4 Training Data Construction 3.5 Training 4 Experiments 4.1 Datasets 4.2 Systems for Comparison 4.3 Evaluation Metrics 4.4 Main Results 5 Experimental Analysis 5.1 Effects of Joint Modeling 5.2 Effects of Initial Translation 5.3 Effects of the Length of Human Typed Characters 5.4 Effects of the Ratio of Retained Context 6 Conclusion References ISTIC's Neural Machine Translation Systems for CCMT' 2023 1 Introduction 2 Data 2.1 Data Size 2.2 Data Preprocessing 3 System 3.1 Systems for Low Resource MT Evaluation Task 3.2 System for Chinese-Centric Multilingual MT Evaluation Task 4 Experiments 4.1 System Environment 4.2 Model and Train 4.3 Experiments Results 5 Conclusion References A Novel Dataset and Benchmark Analysis on Document Image Translation 1 Introduction 2 DITrans 2.1 Fine-Grained Annotations 2.2 Dataset Construction 2.3 Dataset Statistics and Comparison 3 Benchmark 3.1 Layout-Aware Robust DIT Framework 3.2 System Variants 3.3 Benchmark Results 4 Analysis 4.1 Analysis of OCR Noise 4.2 Analysis of Layout Structure 5 Conclusion References Joint Contrastive Learning for Factual Consistency Evaluation of Cross-Lingual Abstract Summarization 1 Introduction 2 Related Work 3 Approach 3.1 Synthetic Training Set 3.2 Human-Annotated Test/Validation Set 3.3 Model 4 Experiments 4.1 Datasets 4.2 Training Details 4.3 Baselines 5 Results and Analysis 5.1 Main Results 5.2 Ablation Study 5.3 Analysis of Contrastive Learning Effectiveness 6 Conclusion References Author Index