Memory-Based Parsing: (7 Natural Language Processing)

Memory-Based Learning (MBL), one of the most influential machine learning paradigms, has been applied with great success to a variety of NLP tasks. This monograph describes the application of MBL to robust parsing. Robust parsing using MBL can provide added functionality for key NLP applications, such as Information Retrieval, Information Extraction, and Question Answering, by facilitating more complex syntactic analysis than is currently available. The text presupposes no prior knowledge of MBL. It provides a comprehensive introduction to the framework and goes on to describe and compare applications of MBL to parsing. Since parsing is not easily characterizable as a classification task, adaptations of standard MBL are necessary. These adaptations can either take the form of a cascade of local classifiers or of a holistic approach for selecting a complete tree.The text provides excellent course material on MBL. It is equally relevant for any researcher concerned with symbolic machine learning, Information Retrieval, Information Extraction, and Question Answering.

Table of Contents:
1. 1. Introduction; 2. 2. Memory-Based Learning; 3. 3. Memory-Based Approaches to Parsing; 4. 4. Data-Oriented Parsing; 5. 5. TuSBL: A Memory-Based Parser; 6. 6. Empirical Evaluation; 7. 7. A Comparison of Memory-Based Approaches to TuSBL; 8. 8. Conclusion and Future Directions; 9. Appendix A. The Stuttgart-Tubingen Tagset; 10. Appendix B The TuBa-D/S Inventory of Syntactic Categories and Grammatical Functions; 11. References; 12. Index of Subjects and Terms

Review :
The book by Sandra Kübler is an important contribution to the area of syntactic parsing in several respects. First, this is the monograph's main point - a memory-based robust parser for German spontaneous speech. A data-driven approach to NLP in its incarnation as an MBL is used for the design of a parser (TueSBL) whose architecture deserves to be looked at by anyone interested in parsing spoken input or using analogy-based methods in Computational Linguistics, or parsing German...Another strong point of the monograph is that the work described in it is clearly placed in the context of other memory-based approaches to parsing. Chapters 3 and 4 give in enough detail to what previous authors have done in this field. The book offers a comprehensive and well-illustrated overview of the area of memory-based parsing, makes all the right methodological points, and describes a system that performs a complex task in a refreshingly simple and smart way. Sandra Kübler's book on memory-based parsing contains a very useful overview of memory-based learning (MBL) as it's been applied to linguistic problems, and it will interest experts for its application to the area of parsing in spoken language dialogues. MBL is at base a classification technique, however, while parsing involves assigning a very specific tree structure to a string of words — a process rather unlikely choosing one of a small number of classes to which an input might belong. Dr. Kuebler's tack is to use MBL to choose a most likely structure, which is then modified to improve it as a structure of the input string. The techniques are implemented are evaluated on the German Verbmobil corpus. This is an important addition to the literature on MBL applied to language, and it is also clearly presented and illustrated. In this monograph, Sandra Kübler proposes a completely novel approach to memory-based parsing which treats memory-based parsing as a classification task over complete trees. Dr. Kübler carefully argues the advantages of her approach over an incremental architecture and presents competitive results for deep parsing of German with both constituency-based and dependency-based evaluations. An important book for researchers interested in data-driven approaches to NLP and for researchers specializing in machine learning. In this enjoyable book, the space of memory-based approaches to parsing is explored, and a highly original new approach is proposed and evaluated. This is a must-read for everyone interested in analogy-based methods applied to parsing and to computational linguistics in general.