As Easy As Vanda, Two, Three: Components for Machine Translation - PowerPoint PPT Presentation

As Easy As Vanda, Two, Three: Components for Machine Translation Based on Formal Grammars Matthias B¨ uchse Theorietag, Prague, 2012-10-04 1 / 18

Outline Basic Principles of Machine Translation State of the Art Vanda: Versatile Components 2 / 18

Outline Basic Principles of Machine Translation State of the Art Vanda: Versatile Components 3 / 18

Goal . . . . . . ich s¨ age ihre ente i saw her duck ich sah, wie sie sich duckte i saw her ducking ich esse spaghetti mit der gabel i eat spaghetti with a fork ich esse spaghetti mit fleischkl¨ oßen i eat spaghetti with meatballs . . . . . . h SL TL 4 / 18

Modelling and Algorithmisation S → � ich s¨ age X , I saw X � X → � ihre Ente , her duck � S ⇒ · · · ⇒ � Ich s¨ age ihre Ente , I saw her duck � modelling select H ⊆ { h | h : SL → TL } e.g., via synchronous grammar 5 / 18

Modelling and Algorithmisation S → � ich s¨ age X , I saw X � X → � ihre Ente , her duck � S ⇒ · · · ⇒ � Ich s¨ age ihre Ente , I saw her duck � modelling select H ⊆ { h | h : SL → TL } e.g., via synchronous grammar program Decoder; ... begin ... end. algorithmise e.g., decoder: given h ∈ H , s ∈ SL compute h ( s ) 5 / 18

Training sentence-aligned bilingual data: parallel corpus 001 Resumption of the session 002 I declare resumed the session of the European Parliament ad- journed on Friday 17 December 1999 , [. . . ] . 001 Wiederaufnahme der Sitzungsperiode 002 Ich erkl¨ are die am Freitag , dem 17. Dezember unterbroche- ne Sitzungsperiode des Europ¨ aischen Parlaments f¨ ur wiederaufge- nommen , [. . . ] . EuroParl corpus, 11 languages, 1.5M sentences each 6 / 18

Training sentence-aligned bilingual data: parallel corpus ↓ apply heuristic, statistical methods 6 / 18

Training sentence-aligned bilingual data: parallel corpus ↓ apply heuristic, statistical methods ↓ grammar rules, weights 6 / 18

Evaluation 1 1 1 2 2 2 3 3 3 . . . . . . . . . SL TL TL ↑ → decoder 7 / 18

Evaluation 1 1 1 score 2 2 2 ↑ 3 3 3 evaluation . . . → e.g. BLEU ← . . . . . . SL TL TL ↑ → decoder 7 / 18

Evaluation 1 1 1 score 2 2 2 ↑ 3 3 3 evaluation . . . → e.g. BLEU ← . . . . . . SL TL TL ↑ → decoder if score > oldscore then publish else perish 7 / 18

Outline Basic Principles of Machine Translation State of the Art Vanda: Versatile Components 8 / 18

Synchronous Context-Free Grammar π 1 : S → � S X, S X � π 2 : S → � X, X � π 3 : X → � yu X 1 you X 2 , have X 2 with X 1 � π 4 : X → � X 1 de X 2 , the X 2 that X 1 � π 5 : X → � X zhiyi, one of X � π 6 : X → � Aozhou, Australia � π 7 : X → � Beihan, North Korea � π 8 : X → � shi, is � π 9 : X → � bangjiao, diplomatic relations � π 10 : X → � shaoshu guojia, few countries � 9 / 18

Derivation � S, S � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � π 4 ⇒ � Aozhou shi X 1 de X 2 zhiyi, Australia is one of the X 2 that X 1 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � π 4 ⇒ � Aozhou shi X 1 de X 2 zhiyi, Australia is one of the X 2 that X 1 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � π 4 ⇒ � Aozhou shi X 1 de X 2 zhiyi, Australia is one of the X 2 that X 1 � π 3 ⇒ � Aozhou shi yu X 1 you X 0 de X 2 zhiyi, Australia is one of the X 2 that have X 0 with X 1 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � π 4 ⇒ � Aozhou shi X 1 de X 2 zhiyi, Australia is one of the X 2 that X 1 � π 3 ⇒ � Aozhou shi yu X 1 you X 0 de X 2 zhiyi, Australia is one of the X 2 that have X 0 with X 1 � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � π 4 ⇒ � Aozhou shi X 1 de X 2 zhiyi, Australia is one of the X 2 that X 1 � π 3 ⇒ � Aozhou shi yu X 1 you X 0 de X 2 zhiyi, Australia is one of the X 2 that have X 0 with X 1 � π 7 ⇒ � Aozhou shi yu Beihan you X 0 de X 2 zhiyi, Australia is one of the X 2 that have X 0 with North Korea � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � π 4 ⇒ � Aozhou shi X 1 de X 2 zhiyi, Australia is one of the X 2 that X 1 � π 3 ⇒ � Aozhou shi yu X 1 you X 0 de X 2 zhiyi, Australia is one of the X 2 that have X 0 with X 1 � π 7 ⇒ � Aozhou shi yu Beihan you X 0 de X 2 zhiyi, Australia is one of the X 2 that have X 0 with North Korea � 10 / 18

Derivation � S, S � π 1 ⇒ � S X, S X � π 1 ⇒ � S X 1 X 2 , S X 1 X 2 � π 2 ⇒ � X 0 X 1 X 2 , X 0 X 1 X 2 � π 6 ⇒ � Aozhou X 1 X 2 , Australia X 1 X 2 � π 8 ⇒ � Aozhou shi X, Australia is X � π 5 ⇒ � Aozhou shi X zhiyi, Australia is one of X � π 4 ⇒ � Aozhou shi X 1 de X 2 zhiyi, Australia is one of the X 2 that X 1 � π 3 ⇒ � Aozhou shi yu X 1 you X 0 de X 2 zhiyi, Australia is one of the X 2 that have X 0 with X 1 � π 7 ⇒ � Aozhou shi yu Beihan you X 0 de X 2 zhiyi, Australia is one of the X 2 that have X 0 with North Korea � π 9 ⇒ � Aozhou shi yu Beihan you bangjiao de X 2 zhiyi, Australia is one of the X 2 that have diplomatic relations with . . . � 10 / 18