pombredanne / juliafoma

Julia NLP with Foma: Finite State Transducer for Morphological Analysis

Julia Foma Interface

(C) 2018 by Damir Cavar

Version: 0.1

This is part of my Julia code and libs for Natural Language Processing (NLP). I am using Julia 1.0 or newer.

This is the beginning of an interface to the Foma library in Julia.

The code here is accompanied by an example morphology in form of a Foma Finite State Transducer.

english.fst

There are a few morphemes and words in this morphology, just for testing purposes.

Make sure that Foma is installed on your machine, and in particular the dynamic libraries. These libraries need to be in your system's library path such that Julia can find them.

This is an extremely fast morphological analyzer. This combination of Julia and Foma FST-based morphological analysis can process more than 300,000 tokens per second on modern Intel i7 CPUs on a single thread.

I will extend the library and functions soon.

Example

For any token that you process using the FST, the output will be of the form:

call+N+Pl
call+V+3P+Sg

This is the output for the input token "calls". The two lines mean that:

• there are two analyses for calls
• the lemma for calls is in both cases call
• the main part of speech is N (noun) and V (verb)
• the noun calls has a morphosyntactic feature +Pl (+plural)
• the verb calls has two morphosyntactic features, that is +3P (third person) and Sg (singular)

Test some more examples using the compiled mini-morphology for English in english.fst.