Syntactic Wordclass Tagging

Part-of-speech (POS) tagging is one of the most popular and thoroughly researched tasks in the field of natural language processing, particularly since it is a prerequisite for a wide variety of more complex tasks. The book Syntactic Wordclass Tagging is a multiauthor collection of articles giving advice on how to use and implement a POS tagger. Part I of the book is entitled "The User's View" and is geared towards novices and researchers who are interested in the POS annotation that taggers produce. Part II, entitled "The Implementer's View," is more technical and is written for researchers who want to understand the advantages of the various computational techniques used for POS tagging; it includes an introductory chapter by Hans van Halteren. After an introductory chapter by Atro Voutilainen, Part I begins with "A Short History of Tagging," also by Voutilainen, which describes some of the notable developments in both data-driven and linguistic approaches to tagging. Then, in "The Use of Tagging," Geoffrey Leech and Nicholas Smith argue that POS tagging is useful for almost every task in corpus linguistics, and has made its way into a number of practical applications as well, such as information retrieval, spelling correction, and machine-aided translation. Leech and Smith also point out that syntactic parsing is arguably the central task of natural language processing, since it is a prerequisite for any kind of semantic analysis of text, and claim that tagging, by being a prerequisite to parsing, is effectively an "entry to the most central area of corpus processing" (p. 27). Jan Cloeren then describes tagsets for wordclass annotation, and discusses the different levels of linguistic details--morphological, syntactic, semantic, discoursal-captured by various tagsets. He also discusses how certain phenomena, such as multiunit tokens, multitoken units, wordclass underspecification, and wordclass ambiguity, present challenges to the design of a tagset. He further describes a proposal by the Text Encoding Initiative (TEI) that specifies how to encode wordclass tags with SGML. "Standards for Tagsets" by Geoffrey Leech and Andrew Wilson discusses the guidelines for POS annotation standards developed by the Expert Advisory Group on Language Engineering Standards, or EAGLES. The chapter describes the obligatory, recommended, and optional attribute-value sets for describing wordclasses across languages and tasks. The intent is that the widely recognized attribute-value pairs (major parts of speech, gender, etc.) would be obligatory or recommended, whereas the task or language-specific attribute-value pairs would be optional. In "Performance of Taggers," Hans van Halteren discusses the relationship between correctness, ambiguity, precision, and recall, which are the most commonly