[ Skip to the content ]

Institute of Formal and Applied Linguistics Wiki


[ Back to the navigation ]

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
user:ptacek:dbmt [2007/05/07 19:10]
ptacek
user:ptacek:dbmt [2007/05/07 19:50] (current)
ptacek
Line 1: Line 1:
 ====== DBMT ====== ====== DBMT ======
-Czech-English Dependency-based Machine Translation+//Czech-English Dependency-based Machine Translation -- Čmejrek, Cuřín, and Havelka 03// 
 +//PCEDT BLEU dtest/etest: 0.1974 0.1704// 
 + 
 +je to Magenta pipeline, jen generovani je rule-based (misto statistikeho tree-to-tree transducing a pak LM) 
 + 
 +na českém prekladu Penn Treebanku 
 + 
 +  - tokenizace a tagging  [Hajic 98] 
 +  - parsing do a_trees [Hajic 98, Charniak 99] 
 +  - afun assigment [ZZ 02] 
 +  - a_tree -> t_tree [Bohmova 01] 
 +  - func assigment C4.5 [ZZ 02] 
 +  - slovnik pomoci GIZA++ [Och and Nay 02] // one most probable translation, 1-2 as 1-1 multiword 
 +  - generator 
 + 
 +====== Generator ====== 
 +dostane TGTS bez tfa, a co koreference :?: 
 + 
 +== 1. determining contextual boundness == 
 +povazuji v CZ od slovesa vlevo jako CB -> definite article 
 +od slovesa vpravo nezapojene -> indefinite article 
 + 
 + 
 +== 2. reordering of constituents == 
 +podle CB se z ACT|PAT|ADDR vyberou Sb 
 +declarative sentence: CB adjuncts + Sb + V + direct/indirect Obj + UB adjuncts 
 + 
 +== 3. generation of verb forms == 
 +pasivum/aktivum se dela podle Sb funktoru 
 + 
 +== 4. insertion of prepositions and articles == 
 +preps: podle české a podle EN nounu 
 + 
 +articles: definite při postmodified NP, premodified by superlative or ordinal num 
 +article prevented: uncountable + proper nouns nebo predetermination by possessive a demonstrative pronouns 
 + 
 +== 5. morphology == 
 +asi ne morpha :!: 
 +hledaji v tabulce  
 +^ word form ^ morphological tag ^ lemma ^ 
 +kdyz nenajdou tak somple rules 
 +taky vokalizace pro indefinite article
  
-je to Magenta pipeline, jen generovani je rule-based misto statistikeho tree-to-tree transducing a pak LM 
  
-  - determining contextual boundness 
-delaji v CZ od slovesa vlevo zapojene, od slovesa vpravo nezapojene -> indefinite article < 
-  - reordering of constituents < 
-  - generation of verb forms < 
-  - insertion of prepositions and articles < 
-  - morphology < 
  

[ Back to the navigation ] [ Back to the content ]