1.3.1 ISO Lexical Markup Framework
1.3.2 Semantic Web Standards
1.4 Chapter Conclusion
2 Linked Lexical Knowledge Bases
2.1 Combining LKBs for Specific Tasks
2.2 Large-scale LLKBs
2.3 Automatic Linking Involving Wordnets
2.4 Manual and Collaborative Linking
2.5 Chapter Conclusion
3.1 Information Integration
3.1.1 Ontology Matching
3.1.2 Database Schema Matching
3.1.3 Graph Matching
3.2 Evaluation Metrics for WSL
3.3 Gloss Similarity-based WSL
3.3.1 Word Overlap
3.3.2 Vector Representations
3.3.3 Personalized PageRank
3.3.4 Additional Remarks
3.4 Graph Structure-based WSL
3.4.1 Wikipedia Category Alignment
3.4.2 Shortest Paths
3.5 Joint Modeling
3.5.1 Machine Learning Approaches
3.5.2 Unsupervised Approaches
3.6 Chapter Conclusion
4 Fundamental Disambiguation Methods
4.1 Disambiguating Textual Units
4.2 Enhanced Disambiguation Using LLKBs
4.2.1 Approaches
4.2.2 Overview of Work in this Area
4.3 Robust Disambiguation Heuristics
4.4 Sense Clustering
4.4.1 Method
4.4.2 Overview of Work in this Area
4.5 Sense-annotated Corpora
4.6 Chapter Conclusion
5 Advanced Disambiguation Methods
5.1 Automatic Knowledge Base Construction
5.2 Distant Supervision
5.2.1 Method
5.2.2 Overview of Work in this Area
5.3 Continuous Vector Space Models of KBs
5.3.1 Method
5.3.2 Overview of Work in this Area
5.4 Chapter Conclusion
6.1 Multilingual Semantic Relatedness
6.2 Computer-aided Translation
6.2.1 Overview of Work in this Area
6.2.2 Illustrative Example
6.3 Chapter Conclusion
7.2 Curation Interfaces
7.3 Resource API’s for Text Processing
7.4 Chapter Conclusion
8.2 Outlook
Foreword
Lexical semantic knowledge is vital for most tasks in natural language processing (NLP). Such knowledge has been captured through two main approaches. The first is the knowledge-based approach, in which human linguistic knowledge is encoded directly in a structured form, resulting in various types of lexical knowledge bases. The second is the corpus-based approach, in which lexical semantic knowledge is learned from corpora and then represented in either explicit or implicit manners.
Historically, the knowledge-based approach preceded the corpus-based one, while the latter has been dominating the center-stage of NLP research in the last decades. Yet, the development and use of lexical knowledge bases (LKBs) continued to be a major thread. An illustration of this fact may be found in the number of citations for the fundamental 1998 WordNet book [Fellbaum, 1998a], over 12,000 at the time of writing (according to Google Scholar), which somewhat exceeds the number of citations for the primary text book on statistical NLP from about the same period [Manning and Schütze, 1999]. Despite the overwhelming success of corpus-based methods, whether supervised or unsupervised, their output may be quite noisy,
