摘 要 i ABSTRACT ii ACKNOWLEDGEMENT iii TABLE OF CONTENTS iv LIST OF TABLES vii LIST OF FIGURES viii Chapter 1. INTRODUCTION 1 1.1 Motivation 1 1.2 Objective 2 1.3 Algorithms for Intelligent Biomedical Summarization 3 1.3.1 Gene Normalization 3 1.3.2 Gene Relation Inquiry from Annotation and Abstracts 4 1.3.3 Full-Text Summarization Using Paragraph Ranking 4 1.4 Organization of Dissertation 5 Chapter 2. RELATED WORKS 7 2.1 Gene Ontology 7 2.2 Performance Measures 8 2.3 Natural Language Processing Tools 9 2.4 Maximum Entropy Model 9 2.5 BioCreAtIvE 10 Chapter 3. GENE NORMALIZATION 12 3.1 Background 12 3.2 Methods 13 3.2.1 Gene mention recognition 14 3.2.2 Matching gene mentions to corresponding identifiers 15 3.2.3 Fuzzy set representation of ambiguous mention 16 3.2.4 Maximum entropy classifiers as membership functions 17 3.2.5 Information fusion for handling ambiguous mentions 18 3.3 Experiment and Results 19 3.3.1 Materials 19 3.3.2 Selection of gene mention recognition tools 20 3.3.3 Evaluation of morphological rules 22 3.3.4 Performance of classifiers 23 3.4 Summery 24 Chapter 4. GENE RELATION SUMMARIZATION 25 4.1 Background 25 4.2 Methods 26 4.2.1 GeneCluster: Measuring semantic similarity of GO term 28 4.2.2 GeneSum: Extracting relations of genes from abstract 30 4.3 Experiments and results 34 4.3.1 Evaluation of GeneCluster 34 4.3.2 Evaluation of GeneSum 36 4.4 Summery 38 Chapter 5. FULL-TEXT SUMMARIZATION 39 5.1 Background 39 5.2 Methods 41 5.2.1 Pre-processing 42 5.2.2 Paragraph Relevance 46 5.2.3 PR-ISR 48 5.2.4 Abstract-related condensed text 49 5.3 Experiments 50 5.3.1 Experimental settings 50 5.3.2 Materials 51 5.3.3 Algorithms for comparison 51 5.3.4 Information overlapping and paragraph importance 52 5.3.5 Assessing agreement with human opinion 53 5.3.6 Evaluation qualities of condensed text 54 5.4 Results and Discussion 55 5.4.1 Analysis of annotation of paragraphs 55 5.4.2 Importance of retrieved paragraphs 56 5.4.3 Information coverage of condensed text 58 5.5 Summery 59 Chapter 6. CONCLUSION AND FUTURE WORKS 60 REFERENCES 62
|