search

rmlockwood / FLExTrans

Machine Translation using FLEx, Apertium, and STAMP
MIT License
10 stars 2 forks source link

[Generate Parses] Improve the module with various features #719

Open rmlockwood opened 2 months ago

rmlockwood commented 2 months ago

The Generate Parses module could use improvements in these areas:

  1. Words selected should be a random sample from the lexicon
  2. Resulting inflections should be given in an order that makes sense (maybe alphabetical by suffixes in the slot)
  3. Clitics and derivational affixes will be added to the generated parses. This should be user selectable (in the Settings) a. A clitic or deriv. affix will only be applied one time to the stem. b. A clitic or deriv. affix will only be applied if the user has defined in the lexicon which category or categories it attaches to. c. The order of things: derivational affixes optionally attached to the stems, then inflectional affixes, then optional clitics d. Never reuse the same derivational affix or clitic in one word.
bbryson commented 1 month ago

I tested the part about "random sample" and "alphabetical order". I like it! Looks really good. (Eventually we hope that the affixes in a template can be ordered in FLEx, and when that happens, it would be nice to use that order. But that hasn't happened yet.)

Are you guys still planning to do #3? Sounds like a good plan.

rmlockwood commented 1 month ago

@bbryson Can you test recent changes in Generate Parses? Clitic and derivational affix generation has been added.

bbryson commented 4 weeks ago

I just downloaded 3.11.1 6 hours ago. I can see the clitics working. I don't see derivation working. I don't know if I didn't set it up right. I have a suffix Adj>Adv, and I asked for Adj stems. It inflected those stems, but didn't add the Adj>Adv suffix to them. Is there a setting that affects whether these get applied? (I don't think there needs to be, given the constraints mentioned above.)

mr-martian commented 4 weeks ago

As currently implemented, you would get the derivational affix if you asked for both adj and adv.