Log in

No account? Create an account
Machine Translation
Technologies, corpora, and evaluation techniques
8th-Jun-2005 11:46 pm
The artificial language Lojban (IPA [ˈloʒban], official full name Lojban: a realization of Loglan) was created by the Logical Language Group in 1987 based on the earlier Loglan, with the intent to make the language more complete, usable, and freely available. It has the ISO 639 language code jbo.

The language itself shares many of the features and goals of Loglan. In particular:

  • The grammar is based on predicate logic, and is capable of expressing complex logical constructs precisely.

  • It has no irregularities or ambiguities in spelling or grammar, so it can be easily parsed by computer.

  • Lojban is designed to be as culturally neutral as possible.

  • It is, nonetheless, simple to learn and use compared to many natural languages.

While the initial goal of the Loglan project was to investigate the Sapir-Whorf Hypothesis, the active Lojban community has additional goals for the language, including

  • General research into linguistics

  • Research in artificial intelligence and machine understanding

  • Improved human-computer communication, storage ontologies, and computer translation of natural language text

  • Use of language as an educational tool

  • Personal creativity

    - Wikipedia, "Lojban"

Has anyone done any work on Lojban as an intermediate language or target language for machine translation?

-William Hsu
20th-Jun-2005 04:35 am (UTC)
Has anyone done any work on Lojban as an intermediate language or target language for machine translation?

It's easy to convert Lojban into other languages (there's a very basic parser (jbofihe) which can spit out semi-comprehensible sentences in English. With a little heuristics, you could probably come up with something almost pleasant to read), so no doubt there's that desire to get things into Lojban. The community is actively working on a multi-language dictionary, so there's that part as well.

I don't know a great deal about machine translation, though know that Lojban may be fairly effective at representing the content of a message. Of course, the hard parts of machine translation would still be hard, but there are things like ERG which hope to help that.

I've been interested in using Lojban as a way to represent my personal knowledge. The difference between Lojban and something like RDF are not as strong as one may think.
6th-Oct-2006 05:34 pm (UTC)
Most of the work in this area has been done by artificial intelligence researcher Ben Goertzel.

Incidentally, the Lojban Livejournal Community is going to blog at least once a day from now on, and welcomes new subscribers.
This page was loaded Apr 22nd 2018, 7:43 pm GMT.