Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
23 Cards in this Set
- Front
- Back
|
translation as decoding |
- Weaver makes an analogy between translation and decoding - even for the most sophisticated codes, decoding is a simpler task than translation - still, the memo inpired reseachers to belive that machine translation might be possible |
|
|
problems in the beginning of machine translation |
- computers weren't very powerful - formal theories of language structure existed, but were not very developed - no computational linguistuc resources |
|
|
a human translator needs 5 different types of information (at least!) |
1. knowledge about the source language 2. knowledge about the target language 3. knowledge about the relationship between the source and target language 4. knowledge about the topic that the text that should be translated is about 5. knowledge about culture, values, traditions and expectation av speakers of source and goal language |
|
|
transfer problems |
no good translation in the goal language for what was said in the source language
these can be at: - concept level - level of style - grammatical level - pragmatic level |
|
|
transfer problem types |
- mismatches - divergences - ambiguities |
|
|
mismatches |
- word-word - word-phrase - phrase-phrase - tempus and aspect - voice |
|
|
divergences |
- thematic - categorical - other types of structural divergence |
|
|
ambiguities |
- lexical - structural - part of speech
|
|
|
what linguistic knowledge is needed to translate? |
-phonological knowledge -- pronunctiation of new words - morphological knowledge -- how do you form words from other words - syntactical knowledge -- how can words be put together to be sentences - semantic knowledge - pragmatic knowledge |
|
|
machine translation systems are traditionally divided into three types of systems |
1. direct translation 2. indirect/transfer translation 3. translation via an interlingua
|
|
|
direct translation |
aka word to word translation/ transformer translation
- translations for words in the source language are looked up in a digital dictionary - very sipmle rules are sometimes used to get some simple regularities correct - goal translation can sometimes be fixed with simple corrections |
|
|
direct translation pros |
- simple to build. might work with closely related language, or when the quality doesn't have to be too good - very robust, is insensitive to bad indata |
|
|
direct translation cons |
- translation isn't very good: no consideration taken of meaning, possible ambiguities etc. - reversing the direction actually requires redoing the entire system |
|
|
indirect/transfer translation |
- simple source language analysis - transfer rules written: map source structures to target structures - target language generation rules: some rules are for cleaning up the output |
|
|
indirect/transfer translation pros |
the translation produced is often grammatical, because more linguistic information is used |
|
|
indirect/transfer translation cons |
- if the source language data contains errors or special words or constructions, the translation will become impossible, the method isn't very robust - our knowledge about syntax isn't very complete, so mistakes arise that weren't conceived of earlier - get alot of syntactic transfer |
|
|
interlingua translation |
- people translate some type of semantic representation. then the semantic representaton is used to build up the goal language structure |
|
|
interlingua pros |
- it's east to add another language - works well with very different languages - certain types of translations are easier because you avoid having to compare very different languages with each other |
|
|
interlingua cons |
- it's very hard to find a suitable representation that contains all the distinctions and values that are needed by each language - the work lies in finding a suitable level of representation , doesn't demand understanding the relationships between the source and goal language (except the semantic differences) |
|
|
Meteo |
- direct translation - three bilingual dictionaries for idioms, place names and weather vocabulary - three processing modules for syntactic analysis of English, the syntactic generation of French and the morphological generation of French - no transfer module because weather reports in English and French have almost the same structure - no need for morpological analysis of English |
|
|
systran |
direct/ some transfer
|
|
|
eurotran |
transfer/indirect |
|
|
rosetta |
interlingua |
