[PDF] [PDF] Computational Linguistics

4 nov 2011 · Why say “Computational Linguistics (CL)” versus “Natural Language Processing ” (NLP)? • Either choose either freely or • Computational 



Previous PDF Next PDF





[PDF] Notes on computational linguistics - UCLA Linguistics

20 fév 2000 · Computational linguistics overlaps to some extent with these other interests, but the goals here are really quite different The notes are very 



[PDF] Computational-Linguisticspdf - Alexander Gelbukh

COMPUTATIONAL LINGUISTICS AND LINGUISTIC MODELS journal Lecture Notes in Computer Science, Springer-Verlag, are marked in boldface



[PDF] Computational Linguistics

4 nov 2011 · Why say “Computational Linguistics (CL)” versus “Natural Language Processing ” (NLP)? • Either choose either freely or • Computational 



[PDF] Introduction to Computational Linguistics - Semantic Scholar

What is Natural Language Processing or Computational Linguistics? • The study of human languages and how they can be represented computationally and 



[PDF] Computational Linguistics and Generative Linguistics: The Triumph

I offer some brief historical notes, and hint at tween computational linguistics ( CL) and general linguistics grammar is like, built into most linguistic theoriza-



[PDF] A Gateway to the Computational Linguistics Curriculum

of computational linguistics is particularly germane for the computer science students Since 2002 the course has been taught three times under the auspices of 



[PDF] Book Review: Foundations of Computational Linguistics: Man

The author notes that while most of the linguistic literature concentrates on the description of internal properties of language, surprisingly little effort is devoted to  



[PDF] Introduction to Computational Linguistics - Brown University

25 oct 2013 · be ignored is a large part of computational linguistics research ) A unigram language model assumes that each word Wi is generated in-



[PDF] The Oxford Handbook Of Computational Linguistics wwwcepunep

Handbook of Computational Linguistics, as one of the most functional sellers here will utterly be accompanied robotics, and critical notes aimed at stimulating

[PDF] computational linguistics ralph grishman

[PDF] computational linguistics reddit

[PDF] computational linguistics stanford

[PDF] computational linguistics uw

[PDF] computational physics

[PDF] compute the fourier series coefficients for the following periodic signals

[PDF] computer animation classes

[PDF] computer animation definition

[PDF] computer animation degree online

[PDF] computer animation jobs

[PDF] computer animation pdf

[PDF] computer animation salary

[PDF] computer animation software

[PDF] computer c programming book pdf

[PDF] computer coding book pdf

Computational Linguistics

Today, we explore this interstice

cartoon from xkcd.com

Computational Linguistics

(aka Natural Language Processing)

Christopher Manning

Ling 1 November 4, 2011

What is Computational Linguistics?

• Getting computers to perform useful tasks involving human languages whether for:

• Enabling human-machine communication • Improving human-human communication • Doing stuff with language data ... email, blogs, etc.

• Examples: • Machine Translation

• Automatic Question Answering • Speech Recognition • Text-to-Speech Synthesis • Text Understanding

CL vs. NLP

• Why say "Computational Linguistics (CL)" versus "Natural Language Processing" (NLP)? • Either choose either freely or ... • Computational Linguistics

• The science of computers dealing with language • Some interest in modeling what people do

• Natural Language Processing • Developing computer systems for processing and understanding human language text

TheVision

Language:theul?mateUI

NLP:Goalsofthefield

• FromtheloXy... • full-onnaturallanguageunderstanding

• par?cipa?oninspokendialogues• open-domainques?onanswering• real-?mebi-direc?onaltransla?on

• ...tothemundane • iden?fyingspam

• categorizingnewsstories• finding&comparingproductinforma?onontheweb• assessingsen?menttowardproducts,brands,stocks,...Predominantinrecentyears

NLPinthecommercialworld

Powerset

Currentmo?va?onsforNLP

• Theexplosionofmachine-readablenaturallanguagetext• Exabytes(10 18

bytes)oftext,doublingeveryyearortwo• Webpages,emails,IMs,SMSs,tweets,docs,PDFs,...• Opportunity - andincreasingnecessity - toextractmeaning

• Media?onofhumaninterac?onsbycomputers • Opportunityforthecomputerinthelooptodomuchmore • Growingroleoflanguageinhuman-computerinterac?on

What'sdrivingNLP?Threetrends:

Furthermo?va?onforCL

Onereasonforstudyinglanguage - andformepersonallythemostcompellingreason - isthatitistemp6ngtoregardlanguage,inthetradi6onalphrase,asa"mirrorofmind".Chomsky,1975Forthesamereason,computa?onallinguis?csisacompellingwaytostudyhumanlanguageacquisi?onandprocessing.Some?mes,thebestwaytounderstandsomethingistobuildamodelofit.WhatIcannotcreate,Idonotunderstand.Feynman,1988

Subfieldsandtasks

Spamdetec?on

SPORTSBUSINESS

YES

Seman?csearch

Whyiscomputa?onallinguis?cshard?

Humanlanguages:

• arehighlyambiguousatalllevels

• arecomplex,withrecursivestructuresandcoreference• subtlyexploitcontexttoconveymeaning• arefuzzyandvague• requirereasoningabouttheworldforunderstanding• arepartofasocialsystem:persuading,insul?ng,amusing,...

OK,whyelseisNLPhard?

Ohsomanyreasons!

non--andardEnglih

Greatjob@jus?nbieber!WereSOOPROUDofwhatyouveaccomplished!Utaughtus2#neversaynever&youyourselfshouldnevergiveupeither -

egmen-a1oniueidiom neologim unfriendretweetbromanceteabagger gardenpa-hen-ence

Themanwhohuntsducksoutonweekends.

Thecokonshirtsaremadefromgrowshere.

-rickyen1-yname ...amuta?onontheforgene... worldknowledge

MaryandSuearesisters.

MaryandSuearemothers.

proody

Ineversaidshestolemymoney.

lexicalpecifici-y

Butthat'swhatmakesitfun!

theNewYork-NewHavenRailroad theNewYork-NewHavenRailroad

Meaningsandexpressions

sodasoXdrinkpopbeverageCoke

Onemeaning,manyexpressions

RussiaincreasingpriceofgasforGeorgia

Consideraseman?csearchapplica?on:

NPNPVPSNPNPPPVPS

photosfromworth1000.com seman?cambiguitysyntac?cambiguity

Ambiguousheadlines

Teacher Strikes Idle Kids

China to Orbit Human on Oct. 15 Juvenile Court to Try Shooting Defendant Clinton Wins on Budget, but More Lies Ahead Local High School Dropouts Cut in Half Police: Crack Found in Man's Buttocks

Parsing

• Parsing is the process of taking a string and a grammar and returning a parse tree or trees for that string

the flight left

Sleft VP Verb NP Det N the flight

Parsing involves search

• As with most everything of interest, parsing involves a search which involves the making of choices

• We'll look at some basic methods to give you an idea of the problem

Top-Down Parsing

• Since we're trying to find trees rooted with an S (Sentence) start with the rules that give us an S.

• Then work your way down from there to the words.

Top-Down Space

S NP VP S VP S NP VP Aux NP VP S NP VP S NP VP Aux S VP NP Aux S NP VP S VP S S

Bottom-Up Parsing

• Of course, we also want trees that cover the input words. So start with trees that link up with the words in the right way.

• Then work your way up from there.

Bottom-Up Space

that flight Book flight that Book flight that Book flight that Book flight that Book flight that Book flight that Book

Control

• We need to keep track of the search space and have a strategy to make choices • We need to systematically explore everything to make sure we find the right parse for a sentence • Which node to try to expand next? • Which grammar rule to use to expand a node?

Top-Down, Depth-First, Left-to-Right Search

Example

Example

[flight] [flight]

Example

flight flight

Efficientparsing

• Thatshouldgiveyouthegeneralideaofhowaparsercanworkbyexploringhypothesessystema?cally • Butreallyweneedtodomuchmoretomake

How to choose between parses?

• Probabilistic methods! • Augment the grammar rules with probabilities • Modify the parser to keep only most probable parses • At the end, return the most probable parse

A statistical scientific revolution

• Computational Linguistics before 1990: • Hand-built parsers, hand-built dialogue systems • High precision, low coverage methods • Computational Linguistics after 1995: • Automatically trained parsers, unsupervised clustering, statistical machine translation • High coverage, low precision methods • Build models exploiting data

Demos!

IfyoumightlikeNLP/CompLing...

• learnJavaorPython(andplaywithJavaNLPorNLTK)

• getsomeexposuretolinguis?cs(LING1,...)• andtologic,probability,sta?s?cs,linearalgebra• studyAIandmachinelearning(CS121,CS221,CS229)• readJurafsky&Mar?norManning&Schütze• Take

• Ling180/CS124:FromLanguagestoInforma?on• Ling284/CS224N:NaturalLanguageProcessing• Ling281/CS224S:SpeechRecogni?on&Synthesis• Ling/CS224U:NaturalLanguageProcessing

Onemorefortheroad

cartoon from qwantz.comquotesdbs_dbs21.pdfusesText_27