Part-of-Speech Tagging and Partial Parsing [chapter]

S. Abney
1997 Text, Speech and Language Technology  
The earliest taggers [35, 51] had large sets of hand-constructed rules for assigning tags on the basis of words' character patterns and on the basis of the tags assigned to preceding or following words, but they had only small lexica, primarily for exceptions to the rules. TAGGIT [35] was used to generate an initial tagging of the Brown corpus, which was then hand-edited. (Thus it provided the data that has since been used to train other taggers [20] .) The tagger described by Garside [56, 34]
more » ... y Garside [56, 34] , CLAWS, was a probabilistic version of TAGGIT, and the DeRose tagger improved on CLAWS by employing dynamic programming. In another line of development, hidden Markov models (HMMs) were imported from speech recognition and applied to tagging, by Bahl and Mercer [9], Derouault and Merialdo [26], and Church [20]. These taggers have come to be standard. Nonetheless, the rule-based line of taggers has continued to be pursued, most notably by Karlsson, Voutilainen, and colleagues [49, 50, 85, 84, 18] and Brill [15, 16] . There have also been efforts at learning parts of speech from word distributions, with application to tagging [76, 77] . Taggers are currently wide-spread and readily available. Those available for free include an HMM tagger implemented at Xerox [23], the Brill tagger, and the Multext tagger [8]. 1 Moreover, taggers have now been developed for a number of different languages. Taggers have been described for Basque [6], Dutch [24], French [18], German [30, 75], Greek [24], Italian [24], Spanish [57], Swedish [13], and Turkish [63], to name a few. Dermatas and Kokkinakis [24] compare taggers for seven different languages. The Multext project [8] is currently developing models to drive their tagger for six languages.
doi:10.1007/978-94-017-1183-8_4 fatcat:nkdxn66w7beita4esb5s55dx44