dr. gv. uma 2

Upload: ilasundaram

Post on 06-Apr-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/2/2019 Dr. GV. Uma 2

    1/50

    NaturalLanguageProcessing

    whatit

    does

    whatisinvolved

    whyisitdifficult

    briefhistor

  • 8/2/2019 Dr. GV. Uma 2

    2/50

    sentencef structuredrepnsofmeaning

    " " . Lisp:

    (query

    (file

    detail

    'date

    "C:/help3.doc"))

    "thelargecatchasedtherat" 1 1 1 1

    L (E1s2 rodent(s2))L chased(s1,s2 )

    "theyoungboyateabadapple" ... ...

  • 8/2/2019 Dr. GV. Uma 2

    3/50

  • 8/2/2019 Dr. GV. Uma 2

    4/50

    whatisinvolved

    ie:symbols

    manipulated

    by

    symbol

    processors

    search&inference

    knowledgerepresentationtechniques

  • 8/2/2019 Dr. GV. Uma 2

    5/50

    whyisitdifficult

    prejudice,politics,

    etc

    ambiguity...

    s ntactic

    semantic

  • 8/2/2019 Dr. GV. Uma 2

    6/50

    examplesentences

    theold

    man

    the

    boats

    mycardrinkspetrol

    IsawtheEiffelTowerflyingtoParis

    heo ened

    the

    door

    with

    the

    ke

    heopenedthedoorwiththesqueakinghinge

    theboykickedtheballunderthetree

    theboy

    kicked

    the

    wall

    under

    the

    tree

    putthebottlesintheboxontheshelfbythedoor

  • 8/2/2019 Dr. GV. Uma 2

    7/50

    (brief)historyoflanguageprocessing

    1950s Russian

    f Englishtranslation

    1956 Chomsky

    1960s Patternmatching

    1970s Parsin

    &some

    KnRe

    1980s Kn&inference

    2000+ quietlypromising

  • 8/2/2019 Dr. GV. Uma 2

    8/50

  • 8/2/2019 Dr. GV. Uma 2

    9/50

    matchin :Student

  • 8/2/2019 Dr. GV. Uma 2

    10/50

    matchin :Elisa

  • 8/2/2019 Dr. GV. Uma 2

    11/50

    amodern inputsentenceapproach

    morphological

    processing

    syntax

    analysis

    lexicon

    semanticanal sis

    grammar

    semantic

    rules

    pragmaticcontextualana ys s

    tar et

    information

    representation

  • 8/2/2019 Dr. GV. Uma 2

    12/50

  • 8/2/2019 Dr. GV. Uma 2

    13/50

    morphologicalprocessing

    notes

    all(?)spoken

    lngs

    exhibit

    morphology

    easiertohandleinwrittenlngsifnoticonic

    somemorphologydescribesinfmbeyondsyntax

    e : roximit Tamil

    Setswana

    etc

    case

  • 8/2/2019 Dr. GV. Uma 2

    14/50

    mor holo exam les

    Noun + Suffix S ntactic case Meanin

    Chennai-ukku dative: destination To Madras

    Chennai-ukku-irundu dative: source From MadrasChennai-le containment In Madras

    Chennai-ai object (formal) Madras

    Fig. 2.2. Suffix Attachment for Noun Cases (Tamil author's spelling)

    Proximity Time Things (inanimate)ear -ppa t s t me: now -n t a t s t ng: t s

    Far a-ppa (that time: then) a-ndtha (that thing: that)

    Question e-ppa (what time: when) e-ndtha (what thing: which)

    Fig. 2.3. Proximity Information as Prefix Tags (Tamil)

    Proximity Cow Student

    Near Speaker kgomo e (this cow) mo-ithuti yo (this student)

    Near Listener k omo e-o that cow mo-ithuti o-o that student

    Far kgomo e-le (that cow) mo-ithuti yo-le (that student)

    Fig. 2.4. Proximity by Demostrative Pronoun Inflection (Setswana)

  • 8/2/2019 Dr. GV. Uma 2

    15/50

    step2

    syntaxanalysis

    objectives: 1 checkfor

    correctness

    2 producephrasestructure

    uses

    parser arulebasedsearchengine

    lexicon dictionaryofwords&theircategories

  • 8/2/2019 Dr. GV. Uma 2

    16/50

    syntaxrules

    partsof

    speech

    cons er

    thecatchasesthemouse

    alllarge

    black

    dogs

    chase

    cats

  • 8/2/2019 Dr. GV. Uma 2

    17/50

    example1

    using

    Lkit

    (build-lexicon

    '((a determiner) (cat noun) (dog noun)

    (build-grammar'((s1 (sentence -> noun-phrase verb-phrase))

    (np (noun-phrase -> determiner noun))

    (vp (verb-phrase -> verb noun-phrase))

    ))

  • 8/2/2019 Dr. GV. Uma 2

    18/50

    example1

    output

    (parse 'sentence '(the dog chased a cat))

    complete-edge 0 5 s1 sentence (the dog ...) nil

    - - -

    Syntax

    (sentence(noun-phrase (determiner the) (noun dog))

    (verb-phrase

    (verb chased)

    (noun-phrase (determiner a) (noun cat))))

    Semantics

  • 8/2/2019 Dr. GV. Uma 2

    19/50

    sowhat?

    wewantmean ng

  • 8/2/2019 Dr. GV. Uma 2

    20/50

    Remember:"theyoungboyateabadapple"

    howcan

    semantics

    be

    encoded

    as

    symbols?

    young/old,happy/sad,good/bad?

    owcansemant cs egenerate

    whatcanbeinferredfromsemantics?

  • 8/2/2019 Dr. GV. Uma 2

    21/50

    Reminder:

    "theyoungboyateabadapple"

  • 8/2/2019 Dr. GV. Uma 2

    22/50

    symbolicrepresentation

    of

    semantics

    (actor (root boy) (id boy#732)

    (tags animate human male)

    (quant specific))

    (action (primitve INGEST))(object (root apple) (id nil)

    (tags physob veg fruit food)

    (qual (phy-state -4))

    (quant non-specific))

  • 8/2/2019 Dr. GV. Uma 2

    23/50

    semanticsin

    lexicon

    asimpleexample(build-lexicon

    '((a det any )

    (chased verb hunts )

    (dog noun canine )

    (the det specific)))

  • 8/2/2019 Dr. GV. Uma 2

    24/50

    semanticsin

    grammar

    rules

    (s1 (sentence -> noun-phrase verb-phrase)

    (actor . noun-phrase)

    action . verb- hrase.action

    (object . verb-phrase.object)

    )

    - -

    (det . noun)

    )

    (vp (verb-phrase -> verb noun-phrase)(action . verb)

    (object . noun-phrase)

    )

  • 8/2/2019 Dr. GV. Uma 2

    25/50

    semantics

    results

    (parse 'sentence '(the dog chased a cat))

    complete-edge 0 5 s1 sentence (the dog...) nil

    s1 sentence -> (noun- hrase verb- hrase)

    Syntax

    (sentence (noun-phrase (det the) (noun dog))

    -

    (noun-phrase (det a) (noun cat))))

    Semantics

    (sentence (actor (specific canine))(action hunts)

    (object (any feline)))

  • 8/2/2019 Dr. GV. Uma 2

    26/50

    semanticsin

    lexicon

    checks

    1

    (a det (sems . any))

    (all det (sems . every))

    . .

    (cats noun (sems . feline) (num . plur))

    (chase verb (sems . hunts) (num . plur))(chases verb (sems . hunts) (num . sing))

    (dog noun (sems . canine) (num . sing))

    (dogs noun (sems . canine) (num . plur))

    (the det (sems . specific))

  • 8/2/2019 Dr. GV. Uma 2

    27/50

    semanticsin

    grammar

    checks

    1

    (s1 (sentence -> noun-phrase verb-phrase)

    (actor . noun-phrase.sems)

    action . verb- hrase.action

    (object . verb-phrase.object)

    ; check number of noun-phrase & verb-phrase

    - .

    = verb-phrase.number)

    numeric-agreement-ok

    numeric-agreement-bad)

    )

  • 8/2/2019 Dr. GV. Uma 2

    28/50

    semantics

    results

    (parse 'sentence '(the dog chases a cat))

    complete-edge 0 5 s1 sentence (the dog...) nil

    s1 sentence -> (noun- hrase verb- hrase)

    Syntax

    (sentence (noun-phrase (det the) (noun dog))

    -

    (noun-phrase (det a) (noun cat))))

    Semantics

    (sentence (actor specific canine)(action . hunts)

    (object any feline)

    numeric-agreement-ok)

  • 8/2/2019 Dr. GV. Uma 2

    29/50

    semantics

    results

    (parse 'sentence '(the dogs chases a cat))

    complete-edge 0 5 s1 sentence (the dog...) nil

    s1 sentence -> (noun- hrase verb- hrase)

    Syntax

    (sentence (noun-phrase (det the) (noun dog))

    -

    (noun-phrase (det a) (noun cat))))

    Semantics

    (sentence (actor specific canine)(action . hunts)

    (object any feline)

    numeric-agreement-bad)

  • 8/2/2019 Dr. GV. Uma 2

    30/50

    semanticsin

    grammar

    checks

    2

    (s1 (sentence -> noun-phrase verb-phrase)

    (fail if noun-phrase.number

    /= verb- hrase.number

    (actor . noun-phrase.sems)

    (action . verb-phrase.action)

    . - .

    )

  • 8/2/2019 Dr. GV. Uma 2

    31/50

    semantics

    results

    (parse 'sentence '(the dog chases a cat))

    Semantics

    sentence actor s ecific canine

    (action . hunts)

    (object any feline))

    (parse 'sentence '(the dogs chases a cat))

    .... failed ....

  • 8/2/2019 Dr. GV. Uma 2

    32/50

    semanticsin

    grammar

    checks

    3

    (s1 (sentence -> noun-phrase verb-phrase)

    (glitch numeric-agreement

    if not noun- hrase.number

    = verb-phrase.number)

    (actor . noun-phrase.sems)

    . - .

    (object . verb-phrase.object)

    )

  • 8/2/2019 Dr. GV. Uma 2

    33/50

    semantics

    results

    (parse 'sentence '(the dogs chases a cat))

    complete-edge 0 5 s1 sentence (the dogs...) nil

    Glitches: numeric-a reement

    s1 sentence -> (noun-phrase verb-phrase)

    Syntax

    -

    (verb-phrase (verb chases)

    (noun-phrase (det a) (noun cat))))

    Semantics(sentence (actor specific canine)

    (action . hunts)

    (object any feline))

  • 8/2/2019 Dr. GV. Uma 2

    34/50

    example2

    lexicon

    (a det any )

    (cat noun feline )

    (chase verb hunts )

    (dog noun canine )

    (the det specific)(black adj (color black))

    (large adj (size 7/10))

    (small adj (size 3/10))

  • 8/2/2019 Dr. GV. Uma 2

    35/50

    example2

    grammar

    (build-grammar

    ' n noun- hrase -> ?det *ad noun

    (if det

    (quantification . det)

    (qualifiers . *.adj)

    (object . noun)

    ))))

  • 8/2/2019 Dr. GV. Uma 2

    36/50

    example2

    results

    (parse 'noun-phrase '(small black dog))

    complete-edge 0 3 np noun-phrase (small...) nil

    n noun- hrase -> (?det *ad noun)

    Syntax

    (noun-phrase (adj small) (adj black) (noun dog))

    (noun-phrase

    (quantification undefined)

    (qualifiers ((size . 3/10)) ((color . black)))(object canine))

  • 8/2/2019 Dr. GV. Uma 2

    37/50

    example2

    results

    small dogs chase the small cats

    and large dogs chase the large cats

    (sentence conjunction

    .

    (object . canine))

    (action . hunts)

    o ect quant . spec c qua s ze .

    (object . feline)))

    ((actor (quant undefined) (qual (size . 7/10))

    (object . canine))(action . hunts)

    (object . feline))))

  • 8/2/2019 Dr. GV. Uma 2

    38/50

    semanticprocessing

    (one

    approach)

    semantic

    rules

    in

    grammarf

    1st

    stage

    case

    frame

    verbformf primitiveactioncaseframe

    disambiguate&

    fill

    additional

    case

    frame

    slots

    checkreferenceswithworldand/ordialog

    dostatementlevelinference

    integratewith

    dialog

    doeventse uencedialo

  • 8/2/2019 Dr. GV. Uma 2

    39/50

    step1:

    produce

    raw

    case

    frame

    verb

    casesthecatchasedtheratinthekitchen

    thecatchasedtheratintothekitchen

    commoncases

    source starttime instrument

    destination endtime beneficiary

    location duration

  • 8/2/2019 Dr. GV. Uma 2

    40/50

    theambiguity

    problem

    eg:

    the

    boy

    kicked

    the

    ball

    under

    the

    tree

    grammarrules

    Sf SPP

    f

    NP f ?det *adj noun

  • 8/2/2019 Dr. GV. Uma 2

    41/50

    exampleframe

    #1

    actor (quant specific)

    (tags animate male human)

    (root boy)

    action (root kick)

    object (root ball)

    (tags manip)

    (posn-relative

    (locator beneath)(object (root tree)

    ... ...

  • 8/2/2019 Dr. GV. Uma 2

    42/50

    exampleframe

    #2

    actor (quant specific)

    (tags animate male human)

    (root boy)

    action (root kick)

    object (root ball)

    (tags manip)

    dest (posn-relative

    (locator beneath)(object (root tree)

    ... ...

  • 8/2/2019 Dr. GV. Uma 2

    43/50

    exampleverb

    form

    #1

    primitive strike

    slots instrument (part-of $actor foot)

    legal start-time, end-time, duration

    instrument, beneficiary, location

    illegal source, dest

  • 8/2/2019 Dr. GV. Uma 2

    44/50

    exampleverb

    form

    #2

    primitive push

    slots instrument (part-of $actor foot)

    legal source, dest, start-time, end-time,

    instr, beneficiary, locatn, duration

  • 8/2/2019 Dr. GV. Uma 2

    45/50

  • 8/2/2019 Dr. GV. Uma 2

    46/50

    integrationwith

    dialog

    dialogshave...

    players(actors)

    props(objects)

    locations

    fromcase

    frames

    themes(derived)

    plans(from

    themes

    and/or

    derived)

  • 8/2/2019 Dr. GV. Uma 2

    47/50

    eventsequence

    setof...

    players(actors)

    props(objects)

    seriesof...

    escapes,exceptions

    &

    alternatives

  • 8/2/2019 Dr. GV. Uma 2

    48/50

    readin rammars etc

    A good source of links & references...

    Computational Analysis of Prepositions http://knol.google.com/k/abdul-baqi-sharaf/computational-analysis-of-prepositions/3hc3uny2z7r41/4#

    if you only plan to read one article...

    Baldwin, T. Kordoni, V and Villavicencio, A. 2009. Prepositions in Applications: ASurvey and Introduction to the Special Issue ". Computational Linguistics 35 (2):119149.

    also...

    Litkowski, Kenneth C. and Orin Hargraves. 2007. SemEval-2007 task 06: Word-sense disambiguation of prepositions. In Proceedings of the 4th InternationalWorkshop on Semantic Evaluations, pages 2429, Prague.

    Disambiguation of Preposition Sense Using Linguistically MotivatedFeatures, Stephen Tratz and Dirk Hovy. Proceedings of the NAACL HLT StudentResearch Workshop and Doctoral Consortium, pages 96100,

    ou er, o ora o, une . c ssoc a on or ompu a ona ngu s cs

  • 8/2/2019 Dr. GV. Uma 2

    49/50

    ,the NLP dictionary: www.cse.unsw.edu.au/~billw/nlpdict.html

    for practical help with building grammars check the following (it is about 10 years

    old but then so is the English language :o)A Grammar Writers Cookbook. Miriam Butt, Tracy Holloway King, Marma-EugeniaNio and Fridirique Segond

    also (for writing larger grammars) it is useful to find a book on grammar for tutorsand/or students of English as a second language.

    for a broad (if a little formal) take on semantics try dipping into...

    Semantics-Oriented Natural Language Processing Mathematical Models andAlgorithms. Vladimir Fomichov A. 2010

  • 8/2/2019 Dr. GV. Uma 2

    50/50

    logic and knowledge representation a guide

    htt : ds ace.dsto.defence. ov.au ds ace bitstream 1947 9996 1 DSTO-TR-2324%20PR.pdf

    representing events for NLP

    htt : www. oo le.co.uk url?sa=t&rct= & =knowled e%20re resentation%20%22representing%20events%22&source=web&cd=6&sqi=2&ved=0CEgQFjAF&url=http%3A%2F%2Fwww.aaai.org%2Focs%2Findex.php%2FFSS%2FFSS10%2Fpaper%2Fdownload%2F2183%2F2819&ei=f6oWT_e7DeKC4gTMpaijBA&usg=AFQjC

    NFYmurwJR9oqfCRBimVprWRK45kew&cad=rja

    semantic networks & frames (2005)

    http://www.cs.bham.ac.uk/~jxb/IAI/w6.pdf

    VERL: An Ontology Framework for Video Events (2005)

    http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1524892