Ideas for Google Summer of Code/Corpus-based lexicalised feature transfer
Make a module that sits somewhere in the Apertium pipeline (somewhere after the lexical selection and before morphological generation) that sets features (e.g. tags) based on a model generated from a corpus. Sometimes we get really inadequate translations even though you'd never hear stuff like that.
One of those things is when we output something as definite when it is never used as definite. One way of dealing with this is a lot of rules and lists in transfer, but those are hard to do. So, how about looking at a corpus for information about some features like definiteness, aspect, evidentiality, impersonal/reflexive pronoun use in Romance languages etc.
- Make a corpus study of one possible feature, the treatment of which could be improved with target-language information.
- Experiment with including a statistical model based on this feature in the Apertium pipeline
- Make a prototype implementation (possibly in python)
- Generalise the prototype to deal with other features
- Come up with an efficient format for storing the model.
- Implement the final program efficiently in C++.
- Make a stream processor (see Apertium stream format) for the output of apertium-transfer (both default/chunk possibilities) that parses character by character.
Frequently asked questions
- none yet, ask us something! :)