Computational Linguistics and Intelligent Text Processing: 20th International Conference, CICLing 2019, La Rochelle, France, April 7–13, 2019, Revised Selected Papers, Part II

Description

The two-volume set LNCS 13451 and 13452 constitutes revised selected papers from the CICLing 2019 conference which took place in La Rochelle, France, April 2019. The total of 95 papers presented in the two volumes was carefully reviewed and selected from 335 submissions. The book also contains 3 invited papers. The papers are organized in the following topical sections: General, Information extraction, Information retrieval, Language modeling, Lexical resources, Machine translation, Morphology, sintax, parsing, Name entity recognition, Semantics and text similarity, Sentiment analysis, Speech processing, Text categorization, Text generation, and Text mining. Preface Organization Contents – Part II Contents – Part I Name Entity Recognition Neural Named Entity Recognition for Kazakh 1 Introduction 2 Related Work 3 Named Entity Features 4 The Neural Networks 4.1 Mapping Words and Tags into Feature Vectors 4.2 Tensor Layer 4.3 Tag Inference 5 Experiments 5.1 Data-Set 5.2 Model Setup 5.3 Results 6 Conclusions References An Empirical Data Selection Schema in Annotation Projection Approach 1 Introduction 2 Related Work 3 Method 3.1 Problems of Previous Method 3.2 Our Method 4 Experiments 4.1 Data Sets and Evaluating Methods 4.2 Results 5 Conclusion References Toponym Identification in Epidemiology Articles – A Deep Learning Approach 1 Introduction 2 Previous Work 3 Our Proposed Model 3.1 Embedding Layer 3.2 Deep Feed Forward Neural Network 4 Experiments and Results 4.1 Effect of Domain Specific Embeddings 4.2 Effect of Linguistic Features 4.3 Effect of Window Size 4.4 Effect of the Loss Function 4.5 Use of Lemmas 5 Discussion 6 Conclusion and Future Work References Named Entity Recognition by Character-Based Word Classification Using a Domain Specific Dictionary 1 Introduction 2 Related Work 3 Baseline Method 4 Proposed Method 5 Experiments 5.1 Datasets 5.2 Methods 5.3 Pre-trained Word Embeddings 5.4 Experimental Results and Discussion 6 Conclusion References Cold Is a Disease and D-cold Is a Drug: Identifying Biological Types of Entities in the Biomedical Domain 1 Introduction 2 Related Work 3 Dataset 4 Approach 4.1 Ontology Creation 4.2 Algorithm: Identify Entity with Its Biological Type 5 Experimental Setup and Results 6 Conclusion References A Hybrid Generative/Discriminative Model for Rapid Prototyping of Domain-Specific Named Entity Recognition 1 Introduction 2 Related Work 2.1 General and Domain-Specific NER 2.2 Types of Supervision in NER 2.3 Unsupervised Word Segmentation and Part-of-Speech Induction 3 Proposed Method 3.1 Task Setting 3.2 Model Overview 3.3 Semi-Markov CRF with a Partially Labeled Corpus 3.4 PYHSMM 3.5 PYHSCRF 4 Experimentals 4.1 Data 4.2 Training Settings 4.3 Baselines 4.4 Results and Discussion 5 Conclusion References Semantics and Text Similarity Spectral Text Similarity Measures 1 Introduction 2 Related Work 3 Similarity Measure/Matrix Norms 4 Document Similarity Measure Based on the Spectral Radius 5 Spectral Norm 6 Application Scenarios 6.1 Market Segmentation 6.2 Translation Matching 7 Evaluation 8 Discussion 9 Supervised Learning 10 Conclusion A Example Contest Answer References A Computational Approach to Measuring the Semantic Divergence of Cognates 1 Introduction 1.1 Related Work 1.2 Contributions 2 The Method 2.1 Cross-Lingual Word Embeddings 2.2 Cross-Language Semantic Divergence 2.3 Detection and Correction of False Friends 3 Conclusions References Triangulation as a Research Method in Experimental Linguistics 1 Introduction 2 Methodology 2.1 Semantic Research and Experiment 2.2 Expert Evaluation Method in Linguistic Experiment 3 Conclusions References Understanding Interpersonal Variations in Word Meanings via Review Target Identification 1 Introduction 2 Related Work 3 Personalized Word Embeddings 3.1 Reviewer-Specific Layers for Personalization 3.2 Reviewer-Universal Layers 3.3 Multi-task Learning of Target Attribute Predictions for Stable Training 3.4 Training 4 Experiments 4.1 Settings 4.2 Overall Results 4.3 Analysis 5 Conclusions References Semantic Roles in VerbNet and FrameNet: Statistical Analysis and Evaluation 1 Introduction 2 VerbNet and FrameNet as Linguistic Resources for Analysis 2.1 VerbNet 2.2 FrameNet 2.3 VerbNet and FrameNet in Comparison 3 Basic Statistical Analysis 4 Advanced Statistical Analysis 4.1 Distribution of Verbs per Class in VN and FN 4.2 Distribution of Roles per Class in VN and FN 4.3 General Analysis and Evaluation 5 Hybrid Role-Scalar Approach 5.1 Hypothesis: Roles Are Not Sufficient for Verb Representation 5.2 Scale Representation 6 Conclusion References Sentiment Analysis Fusing Phonetic Features and Chinese Character Representation for Sentiment Analysis 1 Introduction 2 Related Work 2.1 General Embedding 2.2 Chinese Embedding 3 Model 3.1 Textual Embedding 3.2 Training Visual Features 3.3 Learning Phonetic Features 3.4 Sentence Modeling 3.5 Fusion of Modalities 4 Experiments and Results 4.1 Experimental Setup 4.2 Experiments on Unimodality 4.3 Experiments on Fusion of Modalities 4.4 Validating Phonetic Feature 4.5 Visualization of the Representation 4.6 Who Contributes to the Improvement? 5 Conclusion References Sentiment-Aware Recommendation System for Healthcare Using Social Media 1 Introduction 1.1 Problem Definition 1.2 Motivation 1.3 Contributions 2 Related Works 3 Proposed Framework 3.1 Sentiment Classification 3.2 Top-N Similar Posts Retrieval 3.3 Treatment Suggestion 4 Dataset and Experimental Setup 4.1 Forum Dataset 4.2 Word Embeddings 4.3 Tools Used and Preprocessing 4.4 UMLS Concept Retrieval 4.5 Relevance Judgement for Similar Post Retrieval 5 Experimental Results and Analysis 5.1 Sentiment Classification 5.2 Top-N Similar Post Retrieval 5.3 Treatment Suggestion 6 Conclusion and Future Work References Sentiment Analysis Through Finite State Automata 1 Introduction 2 State of the Art 3 Methodology 3.1 Local Grammars and Finite-State Automata 3.2 Sentita and Its Manually-Built Resources 4 Morphology 5 Syntax 5.1 Opinionated Idioms 5.2 Negation 5.3 Intensification 5.4 Modality 5.5 Comparison 5.6 Other Sentiment Expressions 6 Conclusion References Using Cognitive Learning Method to Analyze Aggression in Social Media Text 1 Introduction 2 Related Work 3 Methodology 3.1 Dataset 3.2 Pre-processing 3.3 Feature Extraction 4 Experiments and Results 4.1 Experimental Setup 4.2 Result 4.3 Discussion and Analysis 5 Conclusion and Future Work References Opinion Spam Detection with Attention-Based LSTM Networks 1 Introduction 2 Related Work 2.1 Opinion Spam Detection 2.2 Deep Learning for Sentiment Analysis 2.3 Attention Mechanisms 3 Methodology 3.1 Attention-Based LSTM Model 4 Experiments 5 Results and Analysis 5.1 All Three-Domain Results 5.2 In-domain Results 5.3 Cross-domain Results 5.4 Comparison with Previous Work 6 Conclusion and Future Work References Multi-task Learning for Detecting Stance in Tweets 1 Introduction 2 Related Work 3 Proposed Approach 3.1 Task Formulation 3.2 Multi-task Learning 3.3 Model Details 4 Experiments 4.1 Dataset 4.2 Training Details 4.3 Baselines 4.4 Evaluation Metrics 4.5 Results 4.6 Ablation Study 4.7 Importance of Regularization 4.8 Effect on Regularization Strength () 4.9 Case-Study and Error Analyses 5 Conclusion References Related Tasks Can Share! A Multi-task Framework for Affective Language 1 Introduction 2 Related Work 3 Proposed Methodology 3.1 Hand-Crafted Features 3.2 Word Embeddings 4 Experiments and Results 4.1 Dataset 4.2 Preprocessing 4.3 Experiments 4.4 Error Analysis 5 Conclusion References Sentiment Analysis and Sentence Classification in Long Book-Search Queries 1 Introduction 2 Related Work 3 User Queries 4 Sentiment Intensity 5 Reviews Language Model 6 Analysing Scores 6.1 Sentiment Intensity, Perplexity and Usefulness Correlation 6.2 Sentiment Intensity, Perplexity and Information Type Correlation 6.3 Graphs Interpretation 7 Conclusion and Future Work References Comparative Analyses of Multilingual Sentiment Analysis Systems for News and Social Media 1 Introduction 1.1 Tasks Description 1.2 Systems Overview 2 Related Work 3 Datasets 3.1 Twitter Datasets 3.2 Targeted Entity Sentiment Datasets 3.3 News Tonality Datasets 4 Evaluation and Results 4.1 Baselines 4.2 Twitter Sentiment Analysis 4.3 Tonality in News 4.4 Targeted Sentiment Analysis 4.5 Error Analysis 5 Conclusion References Sentiment Analysis of Influential Messages for Political Election Forecasting 1 Introduction 2 Related Works 2.1 Sentiment Analysis 2.2 Election Forcasting Approaches 3 Proposed Method 3.1 Data Collection 3.2 Feature Generation 3.3 Influential Classifier Construction 3.4 Election Outcome Prediction Model 4 Results and Findings 4.1 Learning Quality 4.2 Features Quality 4.3 Predicting Election Outcome Quality 5 Conclusion References Basic and Depression Specific Emotions Identification in Tweets: Multi-label Classification Experiments 1 Introduction 1.1 Emotion Modeling 1.2 Multi-label Emotion Mining Approaches 1.3 Problem Transformation Methods 1.4 Algorithmic Adaptation Methods 2 Baseline Models 3 Experiment Models 3.1 A Cost Sensitive RankSVM Model 3.2 A Deep Learning Model 3.3 Loss Function Choices 4 Experiments 4.1 Data Set Preparation 4.2 Feature Sets 4.3 Evaluation Metrics 4.4 Quantifying Imbalance in Labelsets 4.5 Mic/Macro F-Measures 5 Results Analysis 5.1 Performance with Regard to F-Measures 5.2 Performance with Regard to Data Imbalance 5.3 Confusion Matrices 6 Conclusion and Future Work References Generating Word and Document Embeddings for Sentiment Analysis 1 Introduction 2 Related Work 3 Methodology 3.1 Corpus-Based Approach 3.2 Dictionary-Based Approach 3.3 Supervised Contextual 4-Scores 3.4 Combination of the Word Embeddings 3.5 Generating Document Vectors 4 Datasets 5 Experiments 5.1 Preprocessing 5.2 Hyperparameters 5.3 Results 6 Conclusion References Speech Processing Speech Emotion Recognition Using Spontaneous Children's Corpus 1 Introduction 2 Methods 2.1 Data 2.2 Feature Selection 2.3 The i-Vector Paradigm 2.4 Classification Approaches 3 Results 4 Conclusion References Natural Language Interactions in Autonomous Vehicles: Intent Detection and Slot Filling from Passenger Utterances 1 Introduction 1.1 Background 2 Methodology 2.1 Data Collection and Annotation 2.2 Detecting Utterance-Level Intent Types 3 Experiments and Results 3.1 Utterance-Level Intent Detection Experiments 3.2 Slot Filling and Intent Keyword Extraction Experiments 3.3 Speech-to-Text Experiments for AMIE: Training and Testing Models on ASR Outputs 4 Discussion and Conclusion References Audio Summarization with Audio Features and Probability Distribution Divergence 1 Introduction 2 Audio Summarization 3 Probability Distribution Divergence for Audio Summarization 3.1 Audio Signal Pre-processing 3.2 Informativeness Model 3.3 Audio Summary Creation 4 Experimental Evaluation 4.1 Results 5 Conclusions References Multilingual Speech Emotion Recognition on Japanese, English, and German 1 Introduction 2 Methods 2.1 Emotional Speech Data 2.2 Classification Approaches 2.3 Shifted Delta Cepstral (SDC) Coefficients 2.4 Feature Extraction 2.5 Evaluation Measures 3 Results 3.1 Spoken Language Identification Using Emotional Speech Data 3.2 Emotion Recognition Based on a Two-Level Classification Scheme 3.3 Emotion Recognition Using Multilingual Emotion Models 4 Discussion 5 Conclusions References Text Categorization On the Use of Dependencies in Relation Classification of Text with Deep Learning 1 Introduction 2 A Syntactical Word Embedding Taking into Account Dependencies 3 Two Models for Relation Classification Using Syntactical Dependencies 3.1 A CNN Based Relation Classification Model (CNN) 3.2 A Compositional Word Embedding Based Relation Classification Model (FCM) 4 Experiments 4.1 SemEVAL 2010 Corpus 4.2 Employed Word Embeddings 4.3 Experiments with the CNN Model 4.4 Experiments with the FCM Model 4.5 Discussion 5 Conclusion References Multilingual Fake News Detection with Satire 1 Introduction 2 Experimental Framework and Results 2.1 Text Resemblance 2.2 Domain Type Detection 2.3 Classification Results 2.4 Result Analysis 3 Conclusion References Active Learning to Select Unlabeled Examples with Effective Features for Document Classification 1 Introduction 2 Related Works 2.1 Active Learning 2.2 Uncertainty Sampling 3 Proposed Method 4 Experiments 4.1 Data Set 4.2 Experiments on Active Learning 4.3 Experimental Results 5 Conclusion References Effectiveness of Self Normalizing Neural Networks for Text Classification 1 Introduction 2 Related Work 3 Self-Normalizing Neural Networks 3.1 Input Normalization 3.2 Initialization 3.3 SELU Activations 3.4 Alpha Dropout 4 Model 4.1 Word Embeddings are Not Normalized 4.2 ELU Activation as an Alternative to SELU 4.3 Model Architecture 5 Experiments and Datasets 5.1 Datasets 5.2 Baseline Models 5.3 Model Parameters 5.4 Training 6 Results and Discussion 6.1 Results 6.2 Discussion 7 Conclusion References A Study of Text Representations for Hate Speech Detection 1 Introduction 2 Problem Definition 3 Related Work 3.1 Text Representations for Hate Speech 3.2 Classification Approaches 4 Study and Proposed Method 4.1 Text Representations 4.2 Classification Methods 5 Experiments and Results 5.1 Datasets and Experimental Setup 5.2 Results 5.3 Significance Testing 5.4 Discussion 6 Conclusion and Future Work References Comparison of Text Classification Methods Using Deep Learning Neural Networks 1 Introduction 2 Related Work 3 Experimental Evaluation and Analysis 3.1 Non-neural Network Approach 3.2 Analysis of the Experiments 3.3 Comparison Tables 4 Conclusion References Acquisition of Domain-Specific Senses and Its Extrinsic Evaluation Through Text Categorization 1 Introduction 2 Acquisition of Domain-Specific Senses 3 Application to Text Categorization 4 Experiments 4.1 Acquisition of Senses 4.2 Text Categorization 5 Related Work 6 Conclusion References ``News Title Can Be Deceptive'' Title Body Consistency Detection for News Articles Using Text Entailment 1 Introduction 2 Related Work 3 Methodology 3.1 Multilayer Perceptron Model (MLP) 3.2 Convolutional Neural Networks Model (CNN) 3.3 Long Short-Term Memory Model (LSTM) 3.4 Combined CNN and LSTM Model 3.5 Modeling 4 Experiments 4.1 Data 4.2 Experimental Setup 4.3 Results and Discussion 4.4 Error Analysis 5 Conclusion and Future Work References Look Who's Talking: Inferring Speaker Attributes from Personal Longitudinal Dialog 1 Introduction 2 Related Work 3 Conversation Dataset 4 Message Content 5 Groups over Time 6 Conversation Interaction 7 Model 8 Features 9 Experiments 10 Results 11 Conclusion References Computing Classifier-Based Embeddings with the Help of Text2ddc 1 Introduction 2 Related Work 3 Model 3.1 Step 1 and 2: Word Sense Disambiguation 3.2 Step 3: Classifier 3.3 Step 4: Classification Scheme 4 Experiment 4.1 Evaluating text2ddc 4.2 Evaluating CaSe 5 Discussion 5.1 Error Analysis 6 Conclusion References Text Generation HanaNLG: A Flexible Hybrid Approach for Natural Language Generation 1 Introduction 2 Related Work 3 HanaNLG: Our Proposed Approach 3.1 Preprocessing 3.2 Vocabulary Selection 3.3 Sentence Generation 3.4 Sentence Ranking 3.5 Sentence Inflection 4 Experiments 4.1 NLG for Assistive Technologies 4.2 NLG for Opinionated Sentences 5 Evaluation and Results 6 Conclusions References MorphoGen: Full Inflection Generation Using Recurrent Neural Networks 1 Introduction 2 Datasets 3 MorphoGen Architecture 4 Generation Experiments 5 Results 6 Conclusions References EASY: Evaluation System for Summarization 1 Introduction 2 EASY System Design 2.1 Summarization Quality Metrics 2.2 Baselines 3 Implementation Details 3.1 Input Selection 3.2 Metrics 3.3 Baselines 3.4 Correlation of Results 4 Availability and Reproducibility 5 Conclusions References Performance of Evaluation Methods Without Human References for Multi-document Text Summarization 1 Introduction 2 Related Work 2.1 ROUGE-N 2.2 ROUGE-L 2.3 ROUGE-S y ROUGE-SU 3 Evaluation Methods 3.1 Manual Methods 3.2 Automatic Methods 4 Proposed Methodology 5 Obtained Results 5.1 Comparison of the State-of-the-Art Evaluation Methods 6 Conclusions and Future Works References EAGLE: An Enhanced Attention-Based Strategy by Generating Answers from Learning Questions to a Remote Sensing Image 1 Introduction 2 Methodology 2.1 Problem Formulation 2.2 EAGLE: An Enhanced Attention-Based Strategy 2.3 Overall Framework 3 Remote Sensing Question Answering Corpus 3.1 Creation Procedure 3.2 Corpus Statistics 4 Experimental Evaluation 4.1 Models Including Ablative Ones 4.2 Dataset 4.3 Evaluation Metrics 4.4 Implementation Details 4.5 Results and Analysis 5 Related Work 5.1 Visual Question Answering (VQA) with Attention 5.2 Associated Datasets 6 Conclusion References Text Mining Taxonomy-Based Feature Extraction for Document Classification, Clustering and Semantic Analysis 1 Introduction 2 Methodology 2.1 Hierarchy of Word Clusters 2.2 Taxonomy-Augmented Features Given a Set of Predefined Words 2.3 Taxonomy-Augmented Features Given the Hierarchy of Word Clusters 3 Experiments 3.1 Datasets 3.2 Experimental Set-Up 3.3 Experimental Results on Document Classification 3.4 Experimental Results on Document Clustering 3.5 Semantic Analysis 4 Conclusion References Adversarial Training Based Cross-Lingual Emotion Cause Extraction 1 Introduction 2 Related Work 2.1 Emotion Cause Extraction 2.2 Cross-Lingual Emotion Analysis 3 Model 3.1 Task Definition 3.2 Adversarial Training Based Cross-Lingual ECA Model 4 Experiments 4.1 Data Sets 4.2 Experimental Settings and Evaluation Metrics 4.3 Comparisons of Different Methods 4.4 Comparisons of Different Architectures 4.5 Effects of Sampling Methods 4.6 Effects of Different Attention Hops 5 Conclusion and Future Work References Techniques for Jointly Extracting Entities and Relations: A Survey 1 Introduction 2 Problem Definition 3 Motivating Example 4 Overview of Techniques 5 Joint Inference Techniques 6 Joint Models 7 Experimental Evaluation 7.1 Datasets 7.2 Evaluation of End-to-End Relation Extraction 7.3 Domain-Specific Entities and Relations 8 Conclusion References Simple Unsupervised Similarity-Based Aspect Extraction 1 Introduction 2 Background and Definitions 3 Related Work 4 Simple Unsupervised Aspect Extraction 5 Experimental Design 6 Results and Discussion 7 Conclusion References Streaming State Validation Technique for Textual Big Data Using Apache Flink 1 Introduction 2 Preliminaries 2.1 Stateful Stream Processing 2.2 Why Using Apache Flink? 2.3 Apache Flink System 2.4 Core Concepts 3 Design Framework 4 Implementation and Evaluation 4.1 Implementation Setup 4.2 Design of the Implementation 4.3 Experimental Setup 4.4 Results 4.5 Evaluation 4.6 Evaluation Matrices 4.7 Visualization of Results 5 Conclusions and Future Work 5.1 Conclusion 5.2 Future Work References Automatic Extraction of Relevant Keyphrases for the Study of Issue Competition 1 Introduction 2 Related Work 3 Keyphrases Extraction 3.1 Candidate Identification 3.2 Candidate Scoring 3.3 Top n-rank Candidates 4 Experiments 4.1 Evaluation Metric 4.2 Datasets 4.3 Results 5 Key-Phrase Extraction Using Portuguese Parliamentary Debates 5.1 Candidates Selection 5.2 Visualisation 6 Conclusion A Appendix References Author Index