bjascob / pyinflect Goto Github PK

There's a problem with the file encoding 'InflectionRules.py'
I just added this so that setup.py works
# -*- coding: utf-8 -*-

It would be great to be able to extend the infl.csv and overrides.csv files, e.g. a word is missing, so it can just be appended it to the file. So one way this could be possible, if I could specify paths to those files and keep them somewhere accessible and not untouchable in the library folder.
Currently, it's not exactly supported. When I tried to initialize manually the Inflections object, I got an error. But maybe I'm just doing something wrong.

Code:

import spacy
from pyinflect import Inflections

nlp = spacy.load('en_core_web_sm')
inflections = Inflections('path/to/infl.csv', overrides_fn='path/to/overrides.csv')

Error:

ValueError: [E090] Extension 'inflect' already exists on Token. To overwrite the existing extension, set `force=True` on `Token.set_extension`.

Heya,

Like the new version! :)

Some quick suggestions:

1. To make it easier to use, I'd list all the valid POS tags and their meanings in the readme. Some people might not be familiar with PTB tags or what the package can generate.

2. If you have time, I also think it'd be nice if getAllInflections returned a dictionary rather than a list: E.g. {"VB": ("eat"), "VBZ": ("eats"), ...}. That way, we could also see all the forms that are available. It's true you can generally infer this from the position of the tuple in the list, but I think it'd be good to make it explicit.

With this in mind, I've attached the code I wrote previously to convert the infl.txt into JSON format with POS tags: new_agid.zip
I think working from this file should also bypsas most of the other issues you have with overrides.

3. I also thought you were going to make the pos_type argument optional in getAllInflections. So you can specify V, A or N, but also leave it blank if you want to see all options;
   E.g.

> getAllInflections("watch")
{"NN": ("watch"), "VB": ("watch"), ...}