Machine Translation 2 Wikipedia Machine translation, often referred to by the acronym MT, is a sub-field of computational linguistics that investigates the use of computer software to translate text or speech from one natural language to another. An Introduction to Personal Definition Machine Translation MT generally investigates the automatic translation of ”standard” language that can be systematically observed in ordinary communication – e.g. Marcello Federico conversations, news, speeches, business letters, user manuals, etc. –. MT is FBK, Trento - Italy generally not concerned with literature genres, nor creative and sophisticated use of language. For several reasons, such kind of language is simply out of 2016 the scope of MT. 1 For a very interesting introduction to issues related to the translation of literature work see Umberto Eco, ”Experiences in Translation”, U. Toronto Press, 2001. M. Federico MT 2016 M. Federico MT 2016 Outline Introduction to MT 1 3 • Introduction • Applications • Approaches • Brief history • Evaluation • Examples • Text genres • Conclusions References: • P. Koehn, Statistical Machine Translation, Cambridge University Press, 2009. • A. Lopez, Statistical Machine Translation, ACM Computing Surveys, vol. 40, number 3, 2008. • D. Jurafsky and J. H. Martin, Speech and Language Processing, Prentice Hall, 2009. • C. Manning and H. Sch¨ utze, Foundations of Statistical Natural Language Processing, MIT Press, 1999. M. Federico MT 2016 M. Federico MT 2016
Introduction to MT Applications of MT 4 6 Why is machine translation so important? 1 • Information society and production of multilingual content 7 billion people - 193 countries - over 150 o ffi cial languages • Globalization and demand for translation services: 1,000 global companies operating in at least 160 countries • Size of worldwide translation market: 37 billion $ per year ≈ 100 million $ per day • Size of translation industry: 30,000 translation companies, 250,000 translators • MT can improve productivity of human translators: integration of MT with human translation (post-editing) • MT can supply cheap gist translation competitive quality-cost-speed trade-o ff Gist translation for social media 1 Sources: Common Sense Advisory, TAUS M. Federico MT 2016 M. Federico MT 2016 Introduction to MT Applications of MT 5 7 Why is machine translation so di ffi cult? High quality human translation implies: Carrier 12:00 PM • deep and rich understanding of source language and text • sophisticated and creative command of target language Nowadays, feasible goals for machine translation are tasks were: • even approximate translation are helpful (gist translation) • professional translators can take advantage of it (computer assisted translation) • linguistic domain is very focused and limited (apps for travelers) 12:00 PM Carrier In general, di ffi culty of translating depends on how similar the target and source languages are in their vocabulary, grammar, and conceptual structure. Speech translation Apps M. Federico MT 2016 M. Federico MT 2016
Applications of MT Lexical Divergence 8 10 English brother Japanese otooto (younger) oniisan (older) English is Japanese isu (subj animate) aru (subj not animate) English French ıtre (be acquainted with) know conna^ savoir (know a proposition) English French ils (masculine) they elles (feminine) German Berg English hill mountain • some languages make distinctions that other languages don’t • di ffi culty to translate from less specific into more specific information • ?? do language di ff erences enforce di ff erent conceptual structures ?? • ?? do people who speak di ff erent languages think di ff erently ?? 2 Integration in computer assisted translation 2 Watch talk by Lera Boroditsky (U. Stanford), ”How Language Shapes Thought”, fora.tv. M. Federico MT 2016 M. Federico MT 2016 Di ff erences and Similarities of Languages Approaches to MT 9 11 • Universal communicative role of language – names for people, words for talking about women, men, children – every language seems to have nouns and verbs • Di ff erences/similarities across large classes of languages : – Morphology: one vs. many morphemes per words, agglutination vs. fusion – Syntax: Subj-Verb-Obj structure (E) vs. SOV (J) vs. VSO (Irish) – Semantics: mapping of semantic roles and meaning of words e.g. direction/manner of motion indicated by verb/satellite in the bottle floated out (E) → la botella sali´ o flotando (S) • Lexical divergence between languages: – Semantical: there is no corresponding word with the same meaning wall (E) → Wand / Mauer (G, inside/outside) – Syntactical: a word is better translated into another part-of-speech she likes to sing (E,v) → sie singt gerne (D,adv) • Cultural Di ff erences : philosophical argument=is translation possible at all? M. Federico MT 2016 M. Federico MT 2016
Approaches to MT Approaches to MT 12 14 Rough classification according to employed linguistic representations : How is knowledge and linguistic information acquired by the system? • Direct model : translate and re-order single words or n-grams • Hand-crafted : knowledge for analysis, transfer, generation, meaning – basically, no linguistic representation is used representation, or direct translation is manually developed • Transfer model : use explicit knowledge about language di ff erences – most of commercial MT systems fall into this category – analyze lexical and syntactic structure of source sentence – requires lots of human labor and expertise – transfer structures from source to target language – includes: rule-based MT – generate corresponding sentence in the target language • Interlingua model : extract the meaning and express it in the target language • Machine-learned : representations are implemented by mathematical models – analyze lexical, syntactical and semantical structure of source sentence learnable from data, e.g. parallel corpora of human translations – interpret the meaning into a canonical interlingua – much less human e ff ort is needed – generate the target sentence from the interlingua – requires huge amounts of data, the more, the better! – includes: statistical MT and neural MT Notice: required knowledge for the interlingua approach grows linearly with number of languages, rather than to the square. M. Federico MT 2016 M. Federico MT 2016 Approaches to MT Approaches to MT 13 15 Interlingua • Transfer • Interlingua Semantics Semantics • Example-based G s e i • Statistical Word-based s n y e l r a a • Statistical Phrase-based n t A i Transfer o n • Statistical Tree-based Syntax Syntax • Statistical Hierarchical phrase-based • Neural Source Target String String Direct M. Federico MT 2016 M. Federico MT 2016
Transfer Approach Interlingua Approach 16 17 context-free grammar Synchronous context-free grammar • Applied to linguistic domains with a limited number of relations and concepts / NP DT NPB NP DT 1 NPB 2 DT 1 NPB 2 → → NPB JJ NN NPB JJ 1 NN 2 / NN 2 JJ 1 – tourist information, hotel booking, flight reservation, ... → → / NPB NN NPB NN NN → → • Semantics of a sentence can be expressed with predicate argument structure · · · · · · DT the DT the / il → → – I need a twin bed room reservation for tomorrow / JJ north JJ north settentrionale → → – book-room(date=tomorrow,type=single) NN wind NN wind / vento → → · · · · · · • Interlingua language has to be designed carefully (by hand) – for some application formalism similar to SQL language NP NP settentrionale • Processing steps in IBMT: DT NPB DT NPB – extract content from source sentence – map content into SQL like IL format JJ NN NN JJ - generate translation from IL format the north wind il vento settentrionale M. Federico MT 2016 M. Federico MT 2016 Transfer Approach Interlingua Approach 16 18 context-free grammar synchronous context-free grammar • S 3 : I’m arriving on june sixth NP DT NPB NP DT 1 NPB 2 / DT 1 NPB 2 → → • I: give-information+temporal+arrival (who=I, time=(june, md6)) / NPB JJ NN NPB JJ 1 NN 2 NN 2 JJ 1 → → NPB NN NPB NN / NN → → • T: my arrival time is sixth of june · · · · · · DT the DT the / il → → JJ north JJ north / settentrionale • S: no that’s not necessary → → / NN wind NN wind vento → → • I: negate · · · · · · • T: no NP NP settentrionale • S: and i was wondering what you have in the way of rooms available during DT NPB DT NPB that time • I: request-information+availability+room (room-type=question) JJ NN NN JJ • T: what kind of rooms are available? the north wind il vento settentrionale 1 This is a toy example. Working approaches use a very large set of probabilistic and lexicalized rules. 3 S: speech (English), I: Interlingua, T: translation (English) M. Federico MT 2016 M. Federico MT 2016
Recommend
More recommend
