3

I am doing research on machine translation in prolog and I am wondering if there are any suitable dictionaries available in text/xml format for machine translation. With suitable for machine translation I mean that entries are marked with word category like noun, verb, etc and whether the word is male/female or singular/plural etc.

I did some research myself and found some dictionaries like Wiktionary or Omegawiki but these are usually not complete or do not have the required information like word category etc. Specifically I am looking for dictionaries translating single words between English/German/French/Dutch.

The many online translation websites are not suitable for my needs, I would need something that works offline. Are there any standard dictionaries that machine translation researchers use?

2

The big name in computer-based dictionaries is WordNet which groups English lexical items by concept (called synsets). I can be downloaded and used offline. Obviously, this tool is quite powerful and is used extensively a lot in various Computational Linguistic and Natural Language Processing applications. There are even projects to create WordNet equivalents for other languages.

Not surprisingly, some people have tried to map synsets across different languages. While I'm not an expert, a quick Google search revealed multiple projects for linking WordNet to various other ontological hierarchies. Perhaps these would fit your needs.

| improve this answer | |
  • The biggest limitation with WordNet is that as far as I'm aware it's English only. This doesn't prevent it from being a part of the solution though. – hippietrail Sep 10 '13 at 14:27
  • 1
    @hippietrail o.O I link to 4 different projects that attempt to map English WordNet synsets to non-English ontologies (i.e. non-English WordNet synsets). I haven't used any of them, so maybe I misinterpreted their project descriptions? – acattle Sep 11 '13 at 1:31
  • 1
    Yes and that was good but I think that's still going to leave a lot more work than many people would expect, possibly including our OP. – hippietrail Sep 11 '13 at 2:21
  • 1
    @hippietrail Fair enough. One of the big things that synsets don't account for is sentiment. For example, "kill" and "murder" have the same underlying meaning but very different sentiments. What makes translation an art, not a science, is the ability to retain the author's subtext when translating into a new language. – acattle Sep 11 '13 at 3:34
  • Yes I think that's the crux of why translation pairs always seem to work better than interlanguages. – hippietrail Sep 11 '13 at 3:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.