Bài giảng Xử lý ngôn ngữ tự nhiên: Chương 2 - Nguyễn Kiêm Hiếu (ĐH Bách khoa Hà Nội)

Chương 2 Gán nhãn từ loại<br /> PennTreebank<br /> Hidden Markov model<br /> Conditional Random Fields<br /> IT4772 Xử lý ngôn ngữ tự nhiên<br /> Viện CNTT-TT, ĐHBKHN<br /> <br /> Đánh giá<br /> <br /> 2<br /> <br /> Chương 2 Xác định từ loại<br /> PennTreebank<br /> INFORMATION EXTRACTION<br /> <br /> NATURAL LANGUAGE UNDERSTANDING<br /> <br /> END-TO-END<br /> APPLICATIONS<br /> <br /> ●<br /> <br /> Created by University of Pennsylvania<br /> <br /> ●<br /> <br /> Eight-years project: 1989 – 1996<br /> <br /> ●<br /> <br /> 7 millions words of POS tagged texts<br /> <br /> ●<br /> <br /> POS tagset is based on Brown Corpus<br /> <br /> NATURAL LANGUAGE GENERATION<br /> <br /> DATA + LINGUISTICS + MACHINE LEARNING<br /> <br /> 3<br /> <br /> 4<br /> <br /> Penn POS tagset<br /> ●<br /> <br /> CC<br /> He bought a car and a house.<br /> <br /> ●<br /> <br /> CD<br /> Five years later, autocar will be popular.<br /> <br /> ●<br /> <br /> DT<br /> Pierre Vinken will join the board.<br /> <br /> ●<br /> <br /> EX<br /> There is no asbestos in our product now.<br /> <br /> 5<br /> <br /> ●<br /> <br /> IN<br /> <br /> 6<br /> <br /> ●<br /> <br /> The percentage of lung cancer appears to be highest.<br /> <br /> Mr Vinken is chairman of Elsevier N.V.<br /> ●<br /> <br /> ●<br /> <br /> JJ<br /> <br /> MD<br /> US should regulate the class of asbestos.<br /> <br /> Rudolph Agnew was named an executive director.<br /> ●<br /> <br /> JJS<br /> <br /> ●<br /> <br /> JJR<br /> <br /> NN<br /> It’s more than three times the expected number.<br /> <br /> The number of death was higher than expected<br /> <br /> ●<br /> <br /> NNS<br /> Portfolio managers expect further declines in interest rates.<br /> <br /> 7<br /> <br /> 8<br /> <br /> ●<br /> <br /> NNP<br /> <br /> ●<br /> <br /> It expects to obtain regulatory approval.<br /> <br /> Alexis Sanchez joined Manchester United<br /> yesterday.<br /> ●<br /> <br /> ●<br /> <br /> NNPS<br /> <br /> PP$<br /> Shareholders approve its acquisition by Royal Trustco Ltd.<br /> <br /> … the Japan Automobile Dealers’ Association...<br /> ●<br /> <br /> PRP<br /> <br /> ●<br /> <br /> RB<br /> … depends heavily on creativity<br /> <br /> POS<br /> ●<br /> <br /> … at Monday’s auction<br /> <br /> RBR<br /> … worked for the project for more than six years<br /> <br /> 9<br /> <br /> ●<br /> <br /> RBS<br /> <br /> 10<br /> <br /> ●<br /> <br /> … to return home<br /> <br /> the most mundane aspect of its workers<br /> ●<br /> <br /> VB<br /> <br /> TO<br /> <br /> ●<br /> <br /> VBD<br /> the executives joined Mayor William<br /> <br /> He decided to stay<br /> ●<br /> <br /> VBG<br /> … before boarding the buses again<br /> <br /> ●<br /> <br /> VBN<br /> A buffet breakfast was held in the museum<br /> <br /> 11<br /> <br /> 12<br /> <br /> ●<br /> <br /> VBP<br /> <br /> ●<br /> <br /> Plans that give advertisers disscount<br /> ●<br /> <br /> WRB<br /> where employees are assigned lunch partners<br /> <br /> VBZ<br /> The plan is not an attempt<br /> <br /> ●<br /> <br /> WDT<br /> a project that did not include Seymor<br /> <br /> ●<br /> <br /> WP<br /> who couldn’t be reach for comment<br /> 13<br /> <br /> corenlp.run<br /> <br /> 14<br /> <br /> http://45.117.171.213/bknlptool/<br /> <br /> 15<br /> <br /> 16<br /> <br /> Chương 2 Xác định từ loại<br /> Hidden Markov model<br /> <br /> ●<br /> <br /> Chương 2 Xác định từ loại<br /> Hidden Markov model<br /> <br /> Markov model<br /> <br /> ●<br /> <br /> Categorical mixture model<br /> <br /> 17<br /> <br /> 18<br /> <br /> First-order HMM<br /> <br /> Chương 2 Xác định từ loại<br /> Hidden Markov model<br /> <br /> ●<br /> <br /> Transition probability<br /> Pr(xt = NN | xt-1 = DT)<br /> <br /> ●<br /> <br /> DT<br /> <br /> NN<br /> <br /> VBD<br /> <br /> IN<br /> <br /> DT<br /> <br /> NN<br /> <br /> The<br /> <br /> cat<br /> <br /> sat<br /> <br /> on<br /> <br /> the<br /> <br /> mat<br /> <br /> Emission probabitlity<br /> Pr(ot = cat | xt = NN)<br /> <br /> 19<br /> <br /> 20<br /> <br />