Automatic Text Summarization

Juan-Manuel Torres-Moreno is Associate Professor at the Université d'Avignon et des Pays de Vaucluse (UAPV) in France and is head of the research team Natural Language Processing (NLP/TALNE) at the Laboratoire Informatique d'Avignon (LIA). His current research lies within the field of NLP where he is investigating techniques for ATS. His other research interests include sentence compression, information retrieval, machine learning and artificial consciousness.

• Foreword by a Zamora and R Salvador xiFOREWORD BY H SAGGION xvNOTATION xviiIntroduction xixPart 1 Foundations 1Chapter 1 Why Summarize Texts? 31.1 The need for automatic summarization 31.2 Definitions of text summarization 51.3 Categorizing automatic summaries 101.4 Applications of automatic text summarization 131.5 About automatic text summarization 151.6 Conclusion 21Chapter 2 Automatic Text Summarization: Some Important Concepts 232.1 Processes before the process 232.1.1 Sentence-term matrix: the vector space model (VSM) model 262.2 Extraction, abstraction or compression? 282.3 Extraction-based summarization 302.3.1 Surface-level algorithms 312.3.2 Intermediate-level algorithms 332.3.3 Deep parsing algorithms 342.4 Abstract summarization 352.4.1 Frump 352.4.2 Information extraction and abstract generation 382.5 Sentence compression and Fusion 382.5.1 Sentence compression 382.5.2 Multisentence fusion 392.6 The limits of extraction 392.6.1 Cohesion and coherence 402.6.2 The HexTAC experiment 422.7 The evolution of text summarization tasks 432.7.1 Traditional tasks 432.7.2 Current and future problems 452.8 Evaluating summaries 502.9 Conclusion 51Chapter 3 Single-document Summarization 533.1 Historical approaches 533.1.1. Luhn’s Automatic Creation of Literature Abstracts 573.1.2 The Luhn algorithm 593.1.3 Edmundson’s linear combination 613.1.4 Extracts by elimination 643.2 Machine learning approaches 663.2.1 Machine learning parameters 663.3 State-of-the-art approaches 693.4 Latent semantic analysis 733.4.1 Singular value decomposition (SVD) 733.4.2 Sentence weighting by SVD 743.5 Graph-based approaches 763.5.1 PAGERANK and SNA algorithms 773.5.2 Graphs and automatic text summarization 783.5.3 Constructing the graph 793.5.4 Sentence weighting 803.6 DIVTEX: a summarizer based on the divergence of probability distribution 833.7 Cortex 853.7.1 Frequential measures 863.7.2 Hamming measures 873.7.3 Mixed measures 883.7.4 Decision algorithm 893.8 Artex 903.9 Enertex 933.9.1 Spins and neural networks 933.9.2 The textual energy similarity measure 953.9.3. Summarization by extraction and textual energy .. 973.10 Approaches using rhetorical analysis 1023.11 Lexical chains 1073.12 Conclusion 107Chapter 4. Guided Multi-document Summarization 1094.1 Introduction 1094.2 The problems of multidocument summarization 1104.3 DUC/TAC & INEX Tweet Contextualization 1124.4 The taxonomy of MDS methods 1154.4.1 Structure based 1154.4.2 Vector space model based 1164.4.3 Graph based 1174.5 Some multi-document summarization systems and algorithms 1174.5.1 Summons 1184.5.2 Maximal marginal relevance 1194.5.3. A multidocument biography summarization system 1204.5.4 Multi-document ENERTEX 1214.5.5 Mead 1234.5.6 Cats 1264.5.7 SUMUM and SUMMA 1284.5.8 Neo-cortex 1314.6 Update summarization 1344.6.1 Update summarization pilot task at DUC 2007 1344.6.2. Update summarization task at TAC 2008 and 2009 1354.6.3 A minimization-maximization approach 1384.6.4 The ICSI system at TAC 2008 and 2009 1424.6.5 The CBSEAS system at TAC 1454.7 Multidocument summarization by polytopes 1464.8 Redundancy 1484.9 Conclusion 149Part 2 Emerging Systems 151Chapter 5. Multi and Cross-lingual Summarization 1535.1. Multilingualism, the web and automatic summarization 1535.2 Automatic multilingual summarization 1565.3 Mead 1595.4 Summarist 1595.5 Columbia Newsblaster 1615.6 Newsexplorer 1635.7 Google News 1665.8 Caps 1665.9 Automatic cross-lingual summarization 1685.9.1 The quality of machine translation 1695.9.2 A graph-based cross-lingual summarizer 1725.10 Conclusion 177Chapter 6. Source and Domain-specific Summarization 1796.1 Genre, specialized documents and automatic summarization 1796.2 Automatic summarization and organic chemistry 1836.2.1 Yachs2 1836.3 Automatic summarization and biomedicine 1896.3.1 Summterm 1896.3.2 A linguistic-statistical approach 1966.4 Summarizing court decisions 2016.5 Opinion summarization 2046.5.1 CBSEAS at TAC 2008 opinion task 2046.6 Web summarization 2066.6.1 Web page summarization 2066.6.2 OCELOT and the statistical gist 2076.6.3 Multitweet summarization 2116.6.4 Email summarization 2156.7 Conclusion 216Chapter 7 Text Abstracting 2197.1 Abstraction-based automatic summarization 2197.2 Systems using natural language generation 2207.3 An abstract generator using information extraction 2227.4. Guided summarization and a fully abstractive approach 2237.5. Abstraction-based summarization via conceptual graphs 2267.6 Multisentence fusion 2277.6.1 Multisentence fusion via graphs 2287.6.2. Graphs and keyphrase extraction: the TAKAHÉ system 2317.7 Sentence compression 2327.7.1 Symbolic approaches 2357.7.2 Statistical approaches 2367.7.3 A statistical-linguistic approach 2387.8 Conclusion 241Chapter 8 Evaluating Document Summaries 2438.1 How can summaries be evaluated? 2438.2 Extrinsic evaluations 2458.3 Intrinsic evaluations 2468.3.1 The baseline summary 2478.4 TIPSTER SUMMAC evaluation campaigns 2488.4.1 Ad hoc task 2498.4.2 Categorization task 2498.4.3 Question-answering task 2508.5 NTCIR evaluation campaigns 2508.6 DUC/TAC evaluation campaigns 2518.6.1 Manual evaluations 2528.7 CLEF-INEX evaluation campaigns 2548.8. Semi-automatic methods for evaluating summaries .. 2568.8.1 Level of granularity: the sentence 2568.8.2 Level of granularity: words 2578.9 Automatic evaluation via information theory 2638.9.1 Divergence of probability distribution 2658.9.2 Fresa 2668.10 Conclusion 271Conclusion 275Appendix 1 Information Retrieval, Nlp and Ats 281Appendix 2. Automatic Text Summarization Resources 305Bibliography 309Index 343