Self-Training for Biomedical Parsing
Home // Journal // Wall Street Journal
File Size: 170.93 KB
Category: Journal & Wall Street Journal
Source: aclweb.org
is for parser adaptation initially training the parser on one type of data for
which hand-labeled trees are available (eg, Wall Street Journal (M. Mar- cus
Share this info:
Content Inside:
Self-Training for Biomedical Parsing David McClosky and Eugene Charniak Brown Laboratory for Linguistic Information Processing (BLLIP) Brown University Providence, RI 02912 {dmcc|ec}@cs.brown.edu Abstract self-training is so different from the original la- beled data. Section three describes our main Parser self-training is the technique of experiment on standard test data (Clegg and taking an existing parser, parsing extra Shepherd, 2005). Section four looks at some data and then creating a second parser preliminary results we obtained on development by treating the extra data as further training data. Here we apply this tech- data that show in slightly more detail how self- nique to parser adaptation. In partic- training improved the parser. We conclude in ular, we self-train the standard Char- section five. niak/Johnson Penn-Treebank parser us- ing unlabeled biomedical abstracts. This 2 Previous Work achieves an f -score of 84.3% on a stan- dard test set of biomedical abstracts from While self-training has worked in several do- the Genia corpus. This is a 20% error re- mains, the early results on self-training for pars- duction over the best previous result on ing were negative (Steedman et al., 2003; Char- biomedical data (80.2% on the same test set). niak, 1997). However more recent results have shown that it can indeed improve parser perfor- mance (Bacchiani et al., 2006; McClosky et al., 1 Introduction 2006a; McClosky et al., 2006b). Parser self-training is the technique of taking an One possible use for this technique is for existing parser, parsing extra data and then cre- parser adaptation — initially training the parser ating a second parser by treating the extra data on one type of data for which hand-labeled trees as further training data. While for many years it are available (e.g., Wall Street Journal (M. Mar- was thought not to help state-of-the art parsers, cus et al., 1993)) and then self-training on a sec- more recent work has shown otherwise. In this ond type of data in order to adapt the parser paper we apply this technique to parser adap- to the second domain. Interestingly, there is lit- tation. In particular we self-train the standard tle to no data showing that this actually works. Charniak/Johnson Penn-Treebank (C/J) parser Two previous papers would seem to address this using unannotated biomedical data. As is well issue: the work by Bacchiani et al. (2006) and known, biomedical data is hard on parsers be- McClosky et al. (2006b). However, in both cases cause it is so far from more “standard” English. the evidence is equivocal. To our knowledge this is the first application of Bacchiani and Roark train the Roark parser self-training where the gap between the training (Roark, 2001) on trees from the Brown treebank and self-training data is so large. and then self-train and test on data from Wall In section two, we look at previous work. In Street Journal. While they show some improve- particular we note that there is, in fact, very ment (from 75.7% to 80.5% f -score) there are little data on self-training when the corpora for several aspects of this work which leave its re- 101 Proceedings of ACL-08: HLT, Short Papers (Companion Volume), pages 101–104, Columbus, Ohio, USA, June 2008. c 2008 Association for Computational Linguistics
Comments
Related PDF Files
Today Terms
daftar harga suku cadang motor yamaha, harga mobil pickup bekas, harga motor yamaha 135lc, noken as, process control by marlin solution manual, izuzu-phanter, dzogchen center surya das, salicylate free avon, www. yamha jupiterz.com, harga striping supra x 125 R, www.mesin honda modif.com, free e-books, sherlock holmes script pdf, supra X 125 injection, nikon 700d vs canon 5d, harga tie rod toyota corolla, hubberts peak rapidshare, human anatomy and physiology 8th edition download, www.yamaha jupiter z.com, mobil10jt-an, hargamotorbekashondasupra125injection, isuzu panther bekas, Beginning ASP.NET E-Commerce in C: From Novice to Professional fyletype:pdf, The Solid Earth %E2 An Introduction to Global Geophysics torrent, rcvandegraaf, dvr4c on internet, dvr4c on internet, doelstelling anatomische les, introduction to statistical quality control pdf, toko onderdil isuzupather surabaya, HARGA isuzu panther 2009, biology of microorganisms madigan, download, free pdf ebooks, autoradio citroen xsara, JOSE DE JESUS RUBIO AVILA, onderdil racing motor, Buch eos 5d mark II .pdf rapidshare, The Dzogchen Center, rapidshare Bosch Handbook for Automotive Electrics - Automotive Electronics 5th Edition, microbiology an evolving science ifile.it, desenvolvendo websites com php, daftar harga suku cadang toyota, microbiology an evolving science pdf download ifile, Vim 101 Hacks download, microbiology an evolving science pdf download, download Livro Desenvolvendo Websites com PHP, panther 2004, data yamaha jupiter 2009, autoradio xsara, pdf source, matlab simulation of synchronous generator, Harga Isuzu Panther Touring Bekas, sap pdf files, motor suzuki genuine part, pdf e-books strategic planning 2009, Fundamentals of Anatomy and Physiology 6th Edition PDF, supra honda 2000*spesifikasi, daftar harga isuzu panther bekas, harga isuzu panther touring, harga mobil bekas fanther grand taoring 2005, Renault DPO, hand book of print media, the machine that changed the world pdf, Spesifikasi Honda supra fit S, linear optimal control systems kwakernaak rapidshare, about honda supra x 125 cw, spesifikasi honda supra x 125, torrent lasts anatomy, Harga second jupiter z 2010, izusu panther, suku supra x, www.yamaha.com.my, meditations on the occult life pdf, pdf source, accesories isuzu panther, harga onderdil motor, download process control thomas marlin book, deap version 2.1, isuzu panther bekas pick up, onderdil jupiter z, suzuki satria lscm, hargayamaham12009, James Womack Daniel Jones and Daniel Roos called The Machine That Changed the World, suku cadang mazda 626, spesifikasi motor honda new supra x 125 cw, tamiya artikelnummer 58248, suku cadang mazda, introduction to statistical quality control for biology, izuzu panter 2005, daftar harga onderdil mazda, biology of microorganisms ebook, harga onderdil isuzu panther, honda 125 injection, download confessions of an advertising man, PDF SOURCE, harga yamaha jupiter 2009, daftar harga truk isuzu elf bekas, Harga Jupiter 2005, daftarhargasukucadangmobil, spesifikasi mesin supra x 125, Desenvolvendo Websites com PHP download, Rosen TY0710 new york, harga truk, daftar harga truk isuzu elf, m0bil panther, onderdil yamaha, harga panther pick up 90-an, harga panther pick up 90-an, more...