philipw@126.com - PowerPoint PPT Presentation

1 / 28
About This Presentation
Title:

philipw@126.com

Description:

Title: Language and Thought Author: Missy Last modified by: wlf Created Date: 3/10/2005 5:12:19 PM Document presentation format: – PowerPoint PPT presentation

Number of Views:51
Avg rating:3.0/5.0
Slides: 29
Provided by: Miss137
Category:

less

Transcript and Presenter's Notes

Title: philipw@126.com


1
????????
  • ???
  • philipw_at_126.com

2
??
  • ??????/??/??
  • ????/??
  • ?????
  • ??????
  • ?????

3
What is a corpus?
Bodies of natural language material (whole
texts, samples from texts, or sometimes just
unconnected sentences), which are stored in
machine-readable form.
4
Voice data is also corpus!
5
Why use a corpus? Why use electronic text?
  • To study knowledge of language through specimens
    of language use naturally-occurring data ...
  • Accessibility
  • Speed can be analyzed more quickly
  • Accuracy for some tasks, processing e-text is
    more accurate than eye scan

6
??
  • A. by medium printed, electronic text, digitized
    speech, video (e.g. for ASL), mixed
  • C. language variables
  • monolingual vs. multilingual (CHILDES database)
  • original vs. translations (parallel)
  • native speaker vs. learner (e.g. corpora of
    learner compositions)

7
Taxonomies of corpora
  • D. language states
  • synchronic vs. diachronic (e.g. Brown vs.
    Helsinki Diachronic corpus)
  • E. Plain vs. annotated

8
?????
5-10??(??) gt100??(??) 5000??(??) gt1??(???)
9
????????
10
????????(Granger 2003)
11
?????
??????
????
????
? ? ?
? ? ?
? ? ?
? ? ?
12
???????
13
??????
????? ?????????????????? ?????????? ??????????????? ??????????????? ??????????? ???????????? ???????? ??????? ????????????
14
  • ??????????????
  • ??????
  • ??????
  • ??????????
  • ???(WORD)??????????(TEXT)?????
  • ???????????
  • ???????????????????
  • ???????????
  • ???????????
  • ??????????

15
???????
16
?????
17
????
  • ??????,???,(????)
  • SECCL????????,?,??-??-??,
  • ?01-47-01?2001??47??1?????????
  • ???????????????????,????????????(?2001-47)?

18
????
  • ?????
  • ????
  • ????

19
?????
1)ltSPOKENgt Spoken (??) 2)ltTEM4gt
???????? 3)ltGRADE2gt Grade 2 (???) 4)ltYEAR02gt
YEAR 2002 (2002???) 5)ltGROUP01gt Group 01
(?1?) 6)ltTASKTYPE1gt Task Type 1 (????
1) 7)ltSEX1Fgt Sex 1 Female (?? 1 ??),ltSex20gt
Sex 2 Absent (?? 2 ????) 8)ltRANK07gt Rank 07
(????????7)
20
?????
ltSPOKENgt ltTEM 4gt ltGRADE 2gt ltYEAR00gt ltGROUP65gt
ltTASKTYPE 1gt ltSEX 1 Fgt ltSex 2 0gt ltRANK 07gt
21
??????
  • ???????A?B?????
  • ????/??(Self Repetition/Repair)
  • a) ???????????????think??,????think think?
  • b) ???(Long Pause)
  • ???????,???lt,gt??,
  • ????????????,???lt.gt???
  • ?????(0.3?),???? ltgt??,?I think?
  • c) ????(Wrong Pronunciation)
  • ???,???????,???????????????????,?????lt
    gt???very ????????very ltwerigt,
    Loise???noiseltloisegt,Sheep???shipltsheepgt?

22
??????
?????lt gt,????????????, ??,????runned,???? ran
ltrunnedgt? He likes ltlikegt to stay in the hotel.
23
?????
????? ??300???? ???????40??,?????????? ???????
?????? ??18-22? ??????? ??1-4?? ????????????
????9???? ???????100-150???? ???????1600?,?????1
600?? ????100???
24
Annotation of corpora
  • Automatic tagging and manual tagging
  • A. Header mark-up
  • B. Part-of-speech tagging
  • C. Syntactic annotation (parsed corpora)
  • D. Pragmatic annotation
  • E. Rhetorical information
  • F. Discourse structure

25
?????
1)ltWCOMPgt Written Composition (????) 2)ltARGgt
Argumentation(???),ltNARgt Narration(???),ltEXPgt
Exposition(???) 3)ltGRADE3gt Grade 3
(???) 4)ltYEAR02gt YEAR 2002 (2002?) 5)ltTIMEDgt
???? 6)ltUNTIMEDgt ????? 7)ltFYSYgt
???? 8)ltLENGTH300Wgt ????300? 9)ltSCOREgt
?? 10)ltgr-gt grammar error(????) 11) ltsp-gt
spelling error(????) 12) ltmis-gt missing(??)
26
??(POS)????
ltWCOMPgt ltNULLgt ltARGgt ltNULLgt ltGRADE1gt ltNULLgt
ltYR03gt ltNULLgt ltTIMEDgt ltNULLgt ltSCORE?gt ltNULLgt
ltLENGTH243Wgt ltNULLgt ltsgtEducation ltNN1gt is ltVBZgt a
ltAT1gt life-long ltJJgt process ltNN1gt . lt.gt
lt/sgtltsgtMaybe ltRRgt many ltDA2gt people ltNNgt do ltVD0gt
n't ltXXgt realize ltVVIgt it ltPPH1gt , lt,gt but ltCCBgt
it ltPPH1gt is ltVBZgt excatly ltJJgt ture ltNN1gt . lt.gt
lt/sgtltsgtEven ltRRgt when ltCSgt a ltAT1gt baby ltNN1gt is
ltVBZgt still ltRRgt in ltIIgt his ltAPPGEgt ( lt(gt her
ltAPPGEgt ) lt)gt mother ltNN1gt 's ltGEgt stomach ltNN1gt
, lt,gt this ltDD1gt process ltNN1gt has ltVHZgt begun
ltVVNgt . lt.gt lt/sgtltsgtThe ltATgt mother ltNN1gt talks
ltNN2gt to ltIIgt him ltPPHO1gt ( lt(gt her ltAPPGEgt ) lt)gt
, lt,gt touches ltVVZgt her ltAPPGEgt baby ltNN1gt upon
ltIIgt her ltAPPGEgt stomach ltNN1gt , lt,gt and ltCCgt the
ltATgt baby ltNN1gt responds ltVVZgt by ltIIgt making
ltVVGgt sounds ltNN2gt , lt,gt kicking ltVVGgt gentlely
ltRRgt and ltCCgt etc. ltRAgt some ltDDgt experts ltNN2gt
advise ltVV0gt that ltDD1gt music ltNN1gt is ltVBZgt good
ltJJgt for ltIFgt these ltDD2gt babies ltNN2gt . lt.gt
lt/sgtltsgtThat ltDD1gt is ltVBZgt called ltVVNgt " lt"gt
education ltNN1gt before ltCSgt born ltVVNgt . lt.gt "
lt"gt lt/sgtltsgtAfter ltCSgt a ltAT1gt baby ltNN1gt is ltVBZgt
born ltVVNgt , lt,gt he ltPPHS1gt ( lt(gt she ltPPHS1gt )
lt)gt usually ltRRgt will ltVMgt be ltVBIgt taught ltVVNgt
to ltTOgt say ltVVIgt " lt"gt Mom ltNN1gt " lt"gt or ltCCgt "
lt"gt Dad ltNN1gt . lt.gt " lt"gt lt/sgtltsgthe ltPPHS1gt will
ltVMgt also ltRRgt be ltVBIgt lift ltNN1gt upon ltIIgt the
ltATgt floor ltNN1gt to ltTOgt learn ltVVIgt to ltTOgt walk
ltVVIgt . lt.gt lt/sgtltsgtThese ltDD2gt seem ltVV0gt nothing
ltPN1gt special ltJJgt , lt,gt but ltCCBgt everyone ltPN1gt
experince ltVV0gt the ltATgt process ltNN1gt , lt,gt or
ltCCgt how ltRRQgt could ltVMgt all ltDBgt of ltIOgt us
ltPPIO2gt speak ltVV0gt and ltCCgt walk ltVV0gt ? lt?gt
lt/sgtltsgtThat ltDD1gt is ltVBZgt a ltAT1gt kind ltNN1gt of
ltIOgt education ltNN1gt , lt,gt too ltRRgt . lt.gt
lt/sgtltsgtThis ltDD1gt kind ltNN1gt of ltIOgt education
ltNN1gt was ltVBDZgt not ltXXgt paid ltVVNgt enough ltDDgt
attention ltNN1gt , lt,gt until ltCSgt recent ltJJgt
years ltNNT2gt . lt.gt lt/sgtltsgtA ltAT1gt lot ltNN1gt of
ltIOgt researches ltNN2gt have ltVH0gt been ltVBNgt done
ltVDNgt , lt,gt and ltCCgt people ltNNgt find ltVV0gt that
ltCSTgt it ltPPH1gt 's ltVBZgt very ltRGgt important ltJJgt
for ltIFgt a ltAT1gt young ltJJgt baby ltNN1gt to ltTOgt
receive ltVVIgt a ltAT1gt good ltJJgt education ltNN1gt .
lt.gt lt/sgtltsgtAt ltIIgt the ltATgt age ltNN1gt of ltIOgt 7
ltMCgt or ltCCgt 8 ltMCgt , lt,gt children ltNN2gt will
ltVMgt be ltVBIgt sent ltVVNgt to ltIIgt primary ltJJgt
schools ltNN2gt . lt.gt lt/sgtltsgtIt ltPPH1gt 's ltVBZgt "
lt"gt real ltJJgt education ltNN1gt " lt"gt in ltIIgt most
ltDATgt of ltIOgt people ltNNgt 's ltGEgt view ltNN1gt .
lt.gt lt/sgtltsgtIn ltIIgt my ltAPPGEgt opinion ltNN1gt , lt,gt
learning ltVVGgt at ltIIgt school ltNN1gt is ltVBZgt be
ltVBIgt educated ltVVNgt , lt,gt and ltCCgt also ltRRgt a
ltAT1gt kind ltNN1gt of ltIOgt education ltNN1gt . lt.gt
lt/sgtltsgtIn ltIIgt China ltNP1gt , lt,gt students ltNN2gt
are ltVBRgt at ltIIgt school ltNN1gt , lt,gt such ltII21gt
as ltII22gt primary ltJJgt school ltNN1gt , lt,gt junior
ltJJgt middle ltJJgt school ltNN1gt , lt,gt senior ltJJgt
middle ltJJgt school ltNN1gt and ltCCgt university
ltNN1gt . lt.gt lt/sgtltsgtThat ltDD1gt 's ltVBZgt to ltTOgt
say ltVVIgt it ltPPH1gt will ltVMgt take ltVVIgt a ltAT1gt
person ltNN1gt nearly ltRRgt sixteen ltMCgt years
ltNNT2gt to ltTOgt study ltVVIgt at ltIIgt school ltNN1gt .
lt.gt lt/sgtltsgtIt ltPPH1gt 's ltVBZgt a ltAT1gt long ltJJgt
time ltNNT1gt , lt,gt however ltRRgt , lt,gt it ltPPH1gt is
ltVBZgt not ltXXgt enough ltRRgt to ltTOgt learn ltVVIgt .
lt.gt lt/sgtltsgtNowadays ltRTgt , lt,gt it ltPPH1gt 's ltVBZgt
very ltRGgt pouplar ltJJgt for ltIFgt adults ltNN2gt at
ltIIgt work ltNN1gt to ltTOgt learn ltVVIgt . lt.gt
lt/sgtltsgtThey ltPPHS2gt can ltVMgt learn ltVVIgt in ltIIgt
skilling ltVVGgt school ltNN1gt , lt,gt on ltIIgt the
ltATgt internet ltNN1gt and ltCCgt soon ltRRgt . lt.gt
lt/sgtltsgtIt ltPPH1gt proves ltVVZgt that ltDD1gt
education ltNN1gt on ltIIgt adults ltNN2gt is ltVBZgt
still ltRRgt going ltVVGgt on ltRPgt . lt.gt lt/sgtltsgtAs
ltII21gt for ltII22gt old ltJJgt people ltNNgt , lt,gt I
ltPPIS1gt do ltVD0gt n't ltXXgt know ltVVIgt much ltRRgt ,
lt,gt but ltCCBgt I ltPPIS1gt know ltVV0gt there ltEXgt 're
ltVBRgt universities ltNN2gt for ltIFgt the ltATgt old
ltJJgt , lt,gt where ltCSgt they ltPPHS2gt can ltVMgt learn
ltVVIgt writing ltNN1gt and ltCCgt drawing ltNN1gt , lt,gt
music ltNN1gt and ltCCgt a ltAT1gt lot ltNN1gt of ltIOgt
things ltNN2gt . lt.gt lt/sgtltsgtMany ltDA2gt old ltJJgt
people ltNNgt are ltVBRgt fond ltJJgt of ltIOgt morning
ltNNT1gt exercise ltNN1gt , lt,gt and ltCCgt it ltPPH1gt
takes ltVVZgt them ltPPHO2gt a ltAT1gt lot ltNN1gt of
ltIOgt time ltNNT1gt to ltTOgt learn ltVVIgt . lt.gt
lt/sgtltsgtDo ltVD0gt n't ltXXgt you ltPPYgt believe ltVVIgt
that ltDD1gt education ltNN1gt is ltVBZgt a ltAT1gt
life-long ltJJgt process ltNN1gt ? lt?gt lt/sgt
????????!
27
P. Nation S. GRANGER
??? ??????????????????? 9.23-24,
?? ???????? ???? ??ewconference_at_126.com
28
Corpus is the mainstream!-- G. Leech
ltWCOMPgt ltNULLgt ltARGgt ltNULLgt ltGRADE1gt ltNULLgt
ltYR03gt ltNULLgt ltTIMEDgt ltNULLgt ltSCORE?gt ltNULLgt
ltLENGTH243Wgt ltNULLgt ltsgtEducation ltNN1gt is ltVBZgt a
ltAT1gt life-long ltJJgt process ltNN1gt . lt.gt
lt/sgtltsgtMaybe ltRRgt many ltDA2gt people ltNNgt do ltVD0gt
n't ltXXgt realize ltVVIgt it ltPPH1gt , lt,gt but ltCCBgt
it ltPPH1gt is ltVBZgt excatly ltJJgt ture ltNN1gt . lt.gt
lt/sgtltsgtEven ltRRgt when ltCSgt a ltAT1gt baby ltNN1gt is
ltVBZgt still ltRRgt in ltIIgt his ltAPPGEgt ( lt(gt her
ltAPPGEgt ) lt)gt mother ltNN1gt 's ltGEgt stomach ltNN1gt
, lt,gt this ltDD1gt process ltNN1gt has ltVHZgt begun
ltVVNgt . lt.gt lt/sgtltsgtThe ltATgt mother ltNN1gt talks
ltNN2gt to ltIIgt him ltPPHO1gt ( lt(gt her ltAPPGEgt ) lt)gt
, lt,gt touches ltVVZgt her ltAPPGEgt baby ltNN1gt upon
ltIIgt her ltAPPGEgt stomach ltNN1gt , lt,gt and ltCCgt the
ltATgt baby ltNN1gt responds ltVVZgt by ltIIgt making
ltVVGgt sounds ltNN2gt , lt,gt kicking ltVVGgt gentlely
ltRRgt and ltCCgt etc. ltRAgt some ltDDgt experts ltNN2gt
advise ltVV0gt that ltDD1gt music ltNN1gt is ltVBZgt good
ltJJgt for ltIFgt these ltDD2gt babies ltNN2gt . lt.gt
lt/sgtltsgtThat ltDD1gt is ltVBZgt called ltVVNgt " lt"gt
education ltNN1gt before ltCSgt born ltVVNgt . lt.gt "
lt"gt lt/sgtltsgtAfter ltCSgt a ltAT1gt baby ltNN1gt is ltVBZgt
born ltVVNgt , lt,gt he ltPPHS1gt ( lt(gt she ltPPHS1gt )
lt)gt usually ltRRgt will ltVMgt be ltVBIgt taught ltVVNgt
to ltTOgt say ltVVIgt " lt"gt Mom ltNN1gt " lt"gt or ltCCgt "
lt"gt Dad ltNN1gt . lt.gt " lt"gt lt/sgtltsgthe ltPPHS1gt will
ltVMgt also ltRRgt be ltVBIgt lift ltNN1gt upon ltIIgt the
ltATgt floor ltNN1gt to ltTOgt learn ltVVIgt to ltTOgt walk
ltVVIgt . lt.gt lt/sgtltsgtThese ltDD2gt seem ltVV0gt nothing
ltPN1gt special ltJJgt , lt,gt but ltCCBgt everyone ltPN1gt
experince ltVV0gt the ltATgt process ltNN1gt , lt,gt or
ltCCgt how ltRRQgt could ltVMgt all ltDBgt of ltIOgt us
ltPPIO2gt speak ltVV0gt and ltCCgt walk ltVV0gt ? lt?gt
lt/sgtltsgtThat ltDD1gt is ltVBZgt a ltAT1gt kind ltNN1gt of
ltIOgt education ltNN1gt , lt,gt too ltRRgt . lt.gt
lt/sgtltsgtThis ltDD1gt kind ltNN1gt of ltIOgt education
ltNN1gt was ltVBDZgt not ltXXgt paid ltVVNgt enough ltDDgt
attention ltNN1gt , lt,gt until ltCSgt recent ltJJgt
years ltNNT2gt . lt.gt lt/sgtltsgtA ltAT1gt lot ltNN1gt of
ltIOgt researches ltNN2gt have ltVH0gt been ltVBNgt done
ltVDNgt , lt,gt and ltCCgt people ltNNgt find ltVV0gt that
ltCSTgt it ltPPH1gt 's ltVBZgt very ltRGgt important ltJJgt
for ltIFgt a ltAT1gt young ltJJgt baby ltNN1gt to ltTOgt
receive ltVVIgt a ltAT1gt good ltJJgt education ltNN1gt .
lt.gt lt/sgtltsgtAt ltIIgt the ltATgt age ltNN1gt of ltIOgt 7
ltMCgt or ltCCgt 8 ltMCgt , lt,gt children ltNN2gt will
ltVMgt be ltVBIgt sent ltVVNgt to ltIIgt primary ltJJgt
schools ltNN2gt . lt.gt lt/sgtltsgtIt ltPPH1gt 's ltVBZgt "
lt"gt real ltJJgt education ltNN1gt " lt"gt in ltIIgt most
ltDATgt of ltIOgt people ltNNgt 's ltGEgt view ltNN1gt .
lt.gt lt/sgtltsgtIn ltIIgt my ltAPPGEgt opinion ltNN1gt , lt,gt
learning ltVVGgt at ltIIgt school ltNN1gt is ltVBZgt be
ltVBIgt educated ltVVNgt , lt,gt and ltCCgt also ltRRgt a
ltAT1gt kind ltNN1gt of ltIOgt education ltNN1gt . lt.gt
lt/sgtltsgtIn ltIIgt China ltNP1gt , lt,gt students ltNN2gt
are ltVBRgt at ltIIgt school ltNN1gt , lt,gt such ltII21gt
as ltII22gt primary ltJJgt school ltNN1gt , lt,gt junior
ltJJgt middle ltJJgt school ltNN1gt , lt,gt senior ltJJgt
middle ltJJgt school ltNN1gt and ltCCgt university
ltNN1gt . lt.gt lt/sgtltsgtThat ltDD1gt 's ltVBZgt to ltTOgt
say ltVVIgt it ltPPH1gt will ltVMgt take ltVVIgt a ltAT1gt
person ltNN1gt nearly ltRRgt sixteen ltMCgt years
ltNNT2gt to ltTOgt study ltVVIgt at ltIIgt school ltNN1gt .
lt.gt lt/sgtltsgtIt ltPPH1gt 's ltVBZgt a ltAT1gt long ltJJgt
time ltNNT1gt , lt,gt however ltRRgt , lt,gt it ltPPH1gt is
ltVBZgt not ltXXgt enough ltRRgt to ltTOgt learn ltVVIgt .
lt.gt lt/sgtltsgtNowadays ltRTgt , lt,gt it ltPPH1gt 's ltVBZgt
very ltRGgt pouplar ltJJgt for ltIFgt adults ltNN2gt at
ltIIgt work ltNN1gt to ltTOgt learn ltVVIgt . lt.gt
lt/sgtltsgtThey ltPPHS2gt can ltVMgt learn ltVVIgt in ltIIgt
skilling ltVVGgt school ltNN1gt , lt,gt on ltIIgt the
ltATgt internet ltNN1gt and ltCCgt soon ltRRgt . lt.gt
lt/sgtltsgtIt ltPPH1gt proves ltVVZgt that ltDD1gt
education ltNN1gt on ltIIgt adults ltNN2gt is ltVBZgt
still ltRRgt going ltVVGgt on ltRPgt . lt.gt lt/sgtltsgtAs
ltII21gt for ltII22gt old ltJJgt people ltNNgt , lt,gt I
ltPPIS1gt do ltVD0gt n't ltXXgt know ltVVIgt much ltRRgt ,
lt,gt but ltCCBgt I ltPPIS1gt know ltVV0gt there ltEXgt 're
ltVBRgt universities ltNN2gt for ltIFgt the ltATgt old
ltJJgt , lt,gt where ltCSgt they ltPPHS2gt can ltVMgt learn
ltVVIgt writing ltNN1gt and ltCCgt drawing ltNN1gt , lt,gt
music ltNN1gt and ltCCgt a ltAT1gt lot ltNN1gt of ltIOgt
things ltNN2gt . lt.gt lt/sgtltsgtMany ltDA2gt old ltJJgt
people ltNNgt are ltVBRgt fond ltJJgt of ltIOgt morning
ltNNT1gt exercise ltNN1gt , lt,gt and ltCCgt it ltPPH1gt
takes ltVVZgt them ltPPHO2gt a ltAT1gt lot ltNN1gt of
ltIOgt time ltNNT1gt to ltTOgt learn ltVVIgt . lt.gt
lt/sgtltsgtDo ltVD0gt n't ltXXgt you ltPPYgt believe ltVVIgt
that ltDD1gt education ltNN1gt is ltVBZgt a ltAT1gt
life-long ltJJgt process ltNN1gt ? lt?gt lt/sgt
Thank You
Write a Comment
User Comments (0)
About PowerShow.com