A%20Robust%20Shallow%20Parser%20for%20Swedish

1
A Robust Shallow Parser for Swedish

• Ola Knutsson, Johnny Bigert, Viggo Kann
• KTH Nada
• Royal Institute of Technology, Sweden

2
What is robustness?

• Robust against noisy, ill-formed and partial
  natural language data

3
Shallow parsing

• Many NLP-applications do not need full parsing
• Shallow parsing
• A parsing approach
• Pre-processing for full parsing
• A collection of techniques
• Abney - finite state cascades (1991)
• Currently, a lot of attention on ML
• Well suitable for modularization

4
Common modules in a shallow parser

• Tokenizer
• PoS-tagger
• Chunker
• Phrase identifier
• Grammatical function identifier

5
Chunking

• NP Den mycket gamla mannenVC gilladeNP mat

Phrase identification
NP Den AP mycket gamla mannenVC gilladeNP
mat
6
Other parsers for Swedish

• Full parsers UCP (Sågvall Hein)SLE (Gambäck)
• Shallow parsers (phrase structure) Cass-Swe
  (Kokkinakis) Megyesi using ML
• Dependency parsers CG (Birn)FDG (Voutilainen)

7
Granska Text Analyzer (GTA)

• Hand-crafted rules
• Context-free backbone
• Partly object-oriented notation

8
Major phrase categories

• NP Han såg den lilla mannen på bänken
• VC Han har spelat kort hela natten
• PP Han såg spår i sanden
• AP Han ogillade små vita lögner
• ADVP Han vill inte gå på bio.
• INFP Han tycker om att spela

9
Clause boundary identification

• Based on Ejerheds algorithm
• Context-sensitive rules
• Using only PoS information

10
Different kinds of rules

• GTA contains 260 rules
• 200 phrase structure identification rules
• 20 clause boundary identification rules
• 40 disambiguation rules

11
Example rule, NP den lilla bilen

• NPmin_at_
• X(wordcldt wordclrg), den
• X2(wordclab)?,
• Y(wordcljj), lilla
• Z(wordclnn) bilen
• --gt
• action(help, wordclZ.wordcl, pnfundef,
• genderZ.gender, numZ.num,
• specZ.spec, caseZ.case)

12
Clause boundary rule

• cl_at_
• V(sed!sen text!"som" wordcl!sn),
• X((wordclpn pnfsub) (wordclpm casenom)
  
• (wordclnn casenom V.case!gen)
  wordclab),
• ---endleftcontext---,
• Y(wordclkn),
• ---beginrightcontext---,
• Y2(((wordclpn pnfsub) (wordclpm
  casenom)
• (wordclnn casenom) wordclab)
  wordclX.wordcl),
• Z(wordclvb (vbfprs vbfprt vbfimp))
• --gt
• action(help, wordclY.wordcl)

13
The Tetris algorithm

PP till general
PP till general Claes
NP general Claes Olsson
NP boken
NP Fänrik Ax
VC gav
PP till general Claes Olsson
14
The IOB format

• Marcus and Ramshaw 1995
• A phrase/clause tag contains two parts
• Phrase/Clause type, e.g. NP, PP
• One of two tags
• I Inside a phrase/clause
• B Beginning a phrase/clause
• When a word does not belong to a phrase
• 3. O Outside

15
Disagreement error

• De dt.utr/neu.plu.def NPB CLB
• gamla jj.pos.utr/neu.plu.ind/def.nom
  APBNPI CLI
• äppelträdet nn.neu.sin.def.nom NPI CLI
• kan vb.prs.akt.mod VCB CLI
• bli vb.inf.akt.kop VCI CLI
• som kn O CLI
• nya jj.pos.utr/neu.plu.ind/def.nom APB
  CLI
• . mad O CLI

16
Partial input

• Arrangör nn.utr.sin.ind.nom NPB
  CLB
• var vb.prt.akt.kop VCB CLI
• Järfälla pm.gen NPBNPB CLI
• naturskyddsförening nn.utr.sin.ind.nom NPBNPI
  CLI
• där ab ADVPB CLI
• är vb.prs.akt.kop VCB CLI
• medlem nn.utr.sin.ind.nom NPB CLI
• . mad O CLI

17
Noisy data

• Inte ab APB CLB
• så ab ADVPBAPBAPI CLI
• tjck jj.pos.utr.sin.ind.nom APBAPIAPI
  CLI
• som ha O CLB
• det pn.neu.sin.def.sub/obj NPB
  CLI
• ofta ab.pos ADVPB CLI
• står vb.prs.akt VCB CLI
• i pp PPB CLI
• lärobökerna nn.utr.plu.def.nom NPBPPI
  CLI
• mid O CLI

18
Word order violation

• Ympkvisten nn.utr.sin.ind.nom NPB CLB
• inte ab ADVPB CLI
• ska vb.prs.akt.mod VCB CLI
• vara vb.inf.akt.kop VCI CLI
• sådär ab ADVPBAPB CLI
• lång jj.pos.utr.sin.ind.nom APB CLI
• , mid O CLI

19
Evaluation

• Manually corrected output from GTA
• Untuned GTA in the evaluation
• 15 000 words from SUC
• 5 genres

20
F-scores for individual phrase types
Type Accuracy Count
ADVP 81.9 1008
AP 91.3 1332
INFP 81.9 512
NP 91.4 6895
O 94.4 2449
PP 95.3 3886
VC 92.9 2562
Total 88.7
21
F-score for clause boundary identification
Tagger F-score
UNIGRAM 84.2
BRILL 87.3
TNT 88.3
F-score for a baseline identifier was 69.0
22
Applications with GTA

• We are using GTA in
• grammar checking, statistical and rule based
• clustering of medical texts
• CALL-systems
• What do you want to do with GTA?

23
More information

• www.nada.kth.se/theory/projects/xcheck
• Contact Ola Knutsson
• knutsson_at_nada.kth.se