Logo Lanfrica

kscanne/chichewa

Domaine:

natural language processing

Type de record:

software
Créateur:
ksc
Hôte:
NLP resources for Chichewa This repository contains some basic NLP resources for the Chichewa language. Currently, all that's here is a basic lexicon (in the folder `lexicon`) and a simple script for doing morphological generation of verb forms: `scripts/gen.pl`. Also potentially useful is the `makefile` in the `lexicon` directory, which was a quick and dirty attempt to generate a very large Chichewa word list as part of another project. These scripts were written by @kscanne, with linguistic advice from @ceekays. There are better ways of doing all of this. A proper morphological transducer would allow generation and analysis and would be much faster as well. The lexicon files are organized by part of speech. A file named `nA-B.txt` should contain nouns from class A whose plurals are in class B. If B==x, then the noun has no plural form. The script `gen.pl` is designed to work in a pipeline. It reads in lines that look like this: NM+SM+TM+OM^kangaz^AM and outputs all possible surface forms with the given root: akangaza (++a1++^kangaz^a+) achikangaza (++a1++chi^kangaz^a+) aikangaza (++a1++i^kangaz^a+) akakangaza (++a1++ka^kangaz^a+) alikangaza (++a1++li^kangaz^a+) amukangaza (++a1++mu^kangaz^a+) ... The morphological elements NM,SM,TM, etc. are documented in the script. Look at the various makefile targets to see how it's used in practice.