Re: [Ankur-core] Picking between ত্ and ৎ

[Jamil A. <its...@gm...>]

On Sat, Aug 8, 2009 at 6:51 PM, Abu Zaher<za...@gm...> wrote:
> I just had a talk regarding this with Golam Mortaza Bhai, pasting that for
> future references :)
>
> (05:52:23 PM) za...@gm.../HomeC8631CA7: I've mailed you regarding an
> issue betten 'ত্ and 'ৎ', if you get the time, plase feel free to answer
> (05:52:25 PM) Golam Mortuza Hossain: I mean I got
> (05:52:30 PM) za...@gm.../HomeC8631CA7: cool
> (05:52:34 PM) Golam Mortuza Hossain: Please
> (05:52:42 PM) Golam Mortuza Hossain: follow "ৎ"
> (05:53:26 PM) Golam Mortuza Hossain: Khanda-Ta as a separate glyph is now
> Unicode standard
> (05:54:03 PM) Golam Mortuza Hossain: which wasn't the case earlier
> (05:54:41 PM) za...@gm.../HomeC8631CA7: I was following ৎ all this
> time, but came across some sites that have ত্ and the fact that in unicode
> character set ৎ has a comment like this "a dead consonant form of ta,
> without implicit vowel, used in some sequences", that why I thought I
> consult you
> (05:55:48 PM) Golam Mortuza Hossain: the reason for this, earlier there was
> no glyph for "Khanda-Ta" in Unicode
> (05:55:59 PM) za...@gm.../HomeC8631CA7: yeah I know
> (05:57:03 PM) Golam Mortuza Hossain: If you want to make it backward
> compatible then
> (05:57:23 PM) Golam Mortuza Hossain: you could consider mapping "ত্"
> (05:57:31 PM) Golam Mortuza Hossain: to "ৎ"
> (05:57:40 PM) Golam Mortuza Hossain: But it could be tricky
> (05:58:57 PM) za...@gm.../HomeC8631CA7: yeah
> (05:59:07 PM) za...@gm.../HomeC8631CA7: I know, I tried a bit
> (05:59:36 PM) Golam Mortuza Hossain: :-)
> (06:01:17 PM) za...@gm.../HomeC8631CA7: we might need to build a table
> for that, for eg. ত‍্ক - ৎক its always like that isn't it, but we can't map
> like it in উত্তর
> (06:01:36 PM) za...@gm.../HomeC8631CA7: so we might need a to check
> all these :(
> (06:02:32 PM) Golam Mortuza Hossain: If I remember correctly then sometime
> people also
> (06:02:42 PM) Golam Mortuza Hossain: used ZWNJ after Halant
> (06:02:51 PM) za...@gm.../HomeC8631CA7: yeah
> (06:03:03 PM) za...@gm.../HomeC8631CA7: I've seen that too
> (06:03:21 PM) Golam Mortuza Hossain: this case should be easy
> (06:04:30 PM) Golam Mortuza Hossain: also when it appears just before "," ,
> ":", "।", "?", " " etc.
> (06:04:44 PM) za...@gm.../HomeC8631CA7: am alreay running the source
> text through a normalizer right now, becase ড় - ড + nukta, we sometimes get
> text in the complex form and the parser gets confused
> (06:04:54 PM) za...@gm.../HomeC8631CA7: aha
> (06:05:23 PM) Golam Mortuza Hossain: yeah I see
> (06:06:50 PM) za...@gm.../HomeC8631CA7: so you think its do-able
> right?
> (06:07:22 PM) Golam Mortuza Hossain: no
> (06:07:52 PM) za...@gm.../HomeC8631CA7: btw, could I paste this
> conversation in the group just as a reference for the others?
> (06:09:11 PM) Golam Mortuza Hossain: In some cases unambiguous mapping may
> not be possible
> (06:09:16 PM) Golam Mortuza Hossain: Yeah, sure
> (06:13:37 PM) Golam Mortuza Hossain: My suggestion would be handle only "ৎ"
> in the engine.
> (06:15:28 PM) Golam Mortuza Hossain: If needed then mapping should be done
> in text pre-parser.
> (06:16:21 PM) Golam Mortuza Hossain: In the long term "ত্" appearance will
> go away!
> (06:16:30 PM) za...@gm.../HomeC8631CA7: I agree
>

yes, keep working with "ৎ"

> --
> Regards
> Abu Zaher Md. Faridee
>
> http://zaher14.blogspot.com/
> http://sourceforge.net/projects/apertium/
> ---
> Time heals every wound, but time itself is a wound that never heals.
>