Francis Tyers

Or this could be merged into `apertium-posttransfer`.

For the unigram tagger, I think that retraining is fine. For the perceptron and bigram HMM taggers, retraining those `.prob` files might be a challenge. Although honestly, we should get...

@flammie even that doesn't work sometimes ...

@mlforcada Here is a brief summary, based on a small experiment with [`nci-nhi`](https://github.com/apertium/apertium-nci-nhi). * Morphological analysis: Works well * Morphological disambiguation: None of the taggers accepts prefix tags * Separable:...

> I think I filed a bug before when I was workign with a student with bas-btc, perhaps it was just VISL CG 3 had issues with prefixed tags... >...

Not sure, but yeah, I think we'd definitely want to be able to require prefix tags too, although perhaps only right next to the lemma (more or less how the...

@sushain97 it looks like a good idea to rewrite the README to be more descriptive and clear, but in the meantime I've markdownified it as you suggest.

@JRMeyer looks like the problem is because of an ancient g++ version and library dependencies, try removing all traces of g++ 4/5 keeping only g++7 and try again.

This is related: #75

@hectoralos I would do it in posttransfer using the LU and perhaps a 1-2 word context window.