A grammar rule

1
A grammar rule
SVE.GRAM CL.IMP ---------------lt PHR.CATgt
'CL, lt TYPEgt 'IMP, lt WORD.CATgt 'VERB,
lt VERB WORD.CATgt 'VERB, lt INFFgt 'IMP,
lt VERB INFFgt lt INFFgt, lt REG V1.LEMgt
lt LEMgt, lt VERB DIATgt lt DIATgt, ADVANCE,
(ADVERBIALS1// CONTINUE), DO(lt REG V1.LEM
VERBCOMPgt), DO(lt VERB LEX VERBACTIONgt),
ADVANCE, (ADVERBIALS2, ADVANCE// CONTINUE),
(END.OF.SENT, STORE/ CL_COORD) Inactive
filter WORD.CAT or PHR.CAT
2
Transfer rules

• copy feature
• delete feature
• transfer feature
• define feature

3
Copy feature
LABEL TYPE SOURCE lt TYPEgt ?TYPE TARGET lt
TYPEgt ?TYPE TRANSFER
4
Delete feature
LABEL REG SOURCE lt REGgt ANY TARGET ltgt
ltgt TRANSFER
5
Transfer feature
LABEL OBJ.DIR SOURCE lt OBJ.DIRgt
?OBJ.DIR1 TARGET lt OBJ.DIRgt
?OBJ.DIR2 TRANSFER ?OBJ.DIR1 ltgt ?OBJ.DIR2  
6
Define feature
LABEL växellåda SOURCE lt word.catgt noun
lt lexgt växellåda.nn.1 TARGET lt word.catgt
noun lt lexgt gearbox.nn.1 TRANSFER  
7
A contextual lexical rule
i universalstativ ? on universal stand LABEL
i_universalstativ SOURCE lt phr.catgt pp lt
prep lexgt i.pp.1 lt prep word.catgt prep
lt pobjgt ?x1 lt pobj df head lexgt
universalstativ.nn.1 TARGET lt phr.catgt pp
lt prep lexgt on.pp.1 lt prep word.catgt
prep lt pobjgt ?x2 TRANSFER ?x1 ? ?x2  
8
A generation rule
LABEL CL.IMP
X1 ---gt X2 X3 X4
ltX1 PHR.CATgt CL
ltX1 VERBgt ltX2gt
ltX1 TYPEgt IMP
ltX1 OBJ.DIRgt ltX3gt ltX1 SEPgt
ltX4gt
9
A trace
10
Language modules in the new organisation of
MULTRA

• dictionary in a database with different views
• excl. contextually defined lexical transfer rules
• Swedish grammar (excl. morfologi)
• transfer grammar
• incl. contextually defined lexical transfer rules
• generation grammar (excl. morphology)
• temporary generation dictionary (lexemes)

11
The lexical database

• sv-en_LinkLexicon
• en_Inflections
• en_LemmaLexicon
• en_LexemeLexicon
• en_Lexicon
• en_StemLexicon
• sv_Inflections
• sv_LemmaLexicon
• sv_LexemeLexicon
• sv_Lexicon
• sv_StemLexicon

12
Scaling up the dictionary

•  
• Swedish lemmas English lemmas
• befor 369 184
• now 20.883 6.562

13
Scaling up the Swedish-English lexeme dictionary

• word alignment with UWA
• lemmatising the Swedish one-word-units in the
  word links
• heuristic lemmatisation of the English
  counterparts
• basic cleaning
• professionell revision
• genereration of lexemes

14
The MATS process in 10 steps

• extracting text sentences from SGML-dokuments
• tokenisation
• extracting lemmas (SLTL), lexemes (SL TL),
  linguistic codes and valency rules form the db.
• expansion av lemma-, lexeme- och valency codes
• mosy-code ? av-strukture
• parsing

15
Continuation (steps 7-11)

• transfer syntactisk generation
• av-strukture ? mosy-code
• dictionary look-up in db
• finish (capital letter etc.)
• recreate SGML

16
Lexicalistic translation

• Identify (lexical) translation units in the
  source sentence
• Translate each unit separately (considering the
  context)
• Order the result in agreement with a model of the
  target language
• (formulation due to Lars Ahrenberg see further
  the KOMA project)
• See also Beaven, L. John, Shake-and-Bake Machine
  Translation. Coling 92, Nantes, 23-28 Aout 1992.

17
Lexicalistic systems?

• Systran?
• Multra?
• Scots

18
Interlingua translation

• See SN

19
Assignment 2 Hable Con Ella (babelfish)

• Translate the text from English to one of your
  favourite languages using Babelfish
• Make a general quality assessment of the
  translation.
• Specify what transfer rules were applied
• Formulate them in the Multra formalism
• Specify the translation errors that were made and
  discuss how they may be handled.