Dots in words

[phajdan]

Dear Nimeth Laszls,

It is Marco, who is maintaining the English dictionaries extension for OpenOffice and improving the British dictionary (Mozilla + AOO).

I am just writing you because the Hunspell "engine" has an issue (bug) dealing with words that have a dot. For example: "etc." <- It keeps suggesting that the word is wrong.

Thanks!

Kind regards,

Original comment by: marcoagpinto

Original Ticket: hunspell/bugs/262

[krunose]

There are two main scenarios when it comes to abbreviations.

Consider misspellings:

1. dot omitted: 'anthrop' instead 'anthrop.'
2. having dot but misspelling abbreviation: 'anhtrop.' inestad 'anthrop.'

Third scenario is having all of the above at once.

For first to be corrected, dictionary needs to have abbreviations entered with dot, but that will make Hunspell to go in a dot-loop.

If dots are in dic-file, for misspelling 'anhtrop.', Hunspell will suggest 'anthrop.', but after accepting this, the result will be 'anthrop..' and word will still be flagged as misspelled. If suggestion accepted again, result will be 'anthrop...' and so on. This happens with 'husspell -d' and Thunderbird.

Putting abbreviations in dic-file without dot can fix second scenario but it breaks the first because then it can't suggest 'anthrop.' for 'anthrop' (no dot).

Tried SUGSWITHDOTS, nothing changed.

Abbreviations should be put in dictionary with dot to distinguish them from other words. Maybe a hack-fix would be to display dot to user in suggestion but trimming it internally? Would that prevent adding dots to 'misspelled' word? That dot-loop is really nasty.

If anybody is willing to take a look at this, I can provide more informations about current behavior and what's desired. LibreOffice kinda fixed it with collecting abbreviations to a separate file...

Thanks.

[krunose]

Didn't know GitHub will pick up reference like this, I just wanted to make reminder for myself, can it be removed or hide, it's completely irrelevant...

[krunose]

Whatever :D

[dimztimz]

Concerning this issue, will be handled in version 2 so abbreviations are properly handled. I had this in my mind some time ago already.

[Sukender]

Hi there,

I guess my own problem is completely related to this thread.

I'm trying to implement a Lojban spell-check dictionary (Hunspell-based, of course). But this language uses dots ('.') as a normal "letter", not as a sentence separator. There may be dots at the beginning or at the end of a word, such as ".y." and ".i"

• If I enter ".i", then hunspell will say it's okay (expected).
• But if I enter "i", hunspell will propose many suggestions but the right one (ie. ".i").

Adding dot in WORDCHARS or enabling SUGSWITHDOTS doesn't fix anything.
Did I miss something in the affix file? Is is a restriction? Can we workaround this?

Thanks.

Reference:
https://github.com/Sukender/lojban-spell-check
https://github.com/Sukender/lojban-spell-check/tree/master/hunspell

[laszlonemeth]

Dear All,

Word breaking of the command line Hunspell needs the following affix file parameter:

WORDCHARS .

See the tests/base.* test dictionary in the source code for a working example.

Best regards,
Laszlo

[Sukender]

I'm afraid it doesn't solve the issue. This is already what I did...

[krunose]

If I put dot i WORDCHARS, hunspell (command line) will suggest dot if it's not there but only if the rest of abbreviation is spelled correctly: gov > gov. (dic input: gov.) It will not make correct suggestion if abbreviation misspelled as 'gob.' and it's a very small distance from 'gob.' to 'gov.'.

SUGSWITHDOTS doesn't help as it also can't recognize dot as part of abbreviation, it just put dot if dot was in text that's being checked.

Is there a way to improve engine to treat dots in dictionary file differently then dots in text that's checked?

Is there a way to distinguish dot in abbreviations from dot as punctuation as sentence boundary? Dot in dictionary file should not be treated the same way as dot in file being checked. I have feeling that Husnpell is getting all inputs with dots stripped out, is that correct?

How Hunspell is breaking text to words (western languages)?