[pt] Added/improved APs in rule ID:MORRER_PERECER_FALECER

[marcoagpinto]

Heya, @susanaboatto and @p-goulart ,

My improvements in this rule yesterday removed:
https://internal1.languagetool.org/regression-tests/via-http/2024-05-27/pt-BR/
MORRER_PERECER_FALECER[1]: 76 FPs
MORRER_PERECER_FALECER[2]: 3 FPs

Now it will remove tons more.

I guess I will look at the tonight's results and tomorrow remove the "temp_off".

Thanks!

[p-goulart]

LGTM, but I think this rule is a great example of why we need wordnet-style semantic categories.

Where are you getting these exceptions from? Are you using a specific corpus?

[marcoagpinto]

LGTM, but I think this rule is a great example of why we need wordnet-style semantic categories.

Where are you getting these exceptions from? Are you using a specific corpus?

@p-goulart

I used the trick in the wordlist of adding "12 34 56." in front of each line since:
Portuguese (Portugal): 11546 input lines ignored (e.g. not between 10 and 300 chars or at least 4 tokens)

Now I got more 40 000 or 50 000 valid sentences to check, and a lot more than the ones in the nightly results.

Now I can finally make deep testing of the rules, although it takes a very long time.

If next year I am able to buy an i9 computer, I will join both PT-BR and the PT-PT wordlists in one and have hundreds of thousands of more sentences to check.

Anyway, in a shorter reply: for example, if using "gato" removed FPs, then I added "cão" and "cachorro", which is one of the tricks I used in this rule.