Not all languages have grammars

[kormanowsky]

Hi, I've just installed FreeLing v4.2 and noticed that not all languages have grammar files needed to run analyzer script to obtain dependencies from a text. I have a file named text.txt which has some Russian text in it and I want analyze to show me the results of text parsing. I run it as follows:
analyze --output json --lang ru --inplv text --outlv dep -f ru.cfg < text.txt
But it fails with an error:
GRAMMAR: Error opening file '/usr/local/share/freeling/ru/chunker/grammar-chunk.dat'

Is this project still in development? Will grammar files be added for all the languages?

[kormanowsky]

@lluisp

[lluisp]

Given the new NLP paradigm that LLMs provide, FreeLing is not under much development, althought it is still maintained.

In any case, all linguistic data are contributed by users.
To have a dependency parser for Russian, you'd need to either develop a grammar for the rule-based parser (not recommended, it is very time consuming with no specially good performance results), or to train the LSTM parser module with a dependency treebank.

There are several dependency treebanks for Russian, based on Universal Dependencies, which could be used to train the parser.
Unfortunately, Russian PoS tagger is not based on UD, but on EAGLES tagset, so the morphological dictionary and the PoS tagger should be re-created and retrained from scratch in order to feed the proper tagset to the parser.

If you are willing to work on this direction and contribute, I'll be glad to guide you in the process.