Representing P most frequent words without n-grams about fasttext HOT 5 CLOSED

I also couldn't find any equivalent to the 'Subword Information' paper's P in fastText.

But rather than modifying computeNgrams source, it should be enough to just provide an extreme -minn or -maxn parameter.

A minn larger than maxn should mean line

from fasttext.

Am I allowed to give a comment
On Aug 8, 2016 12:01 AM, "Gordon Mohr" [email protected] wrote:

I also couldn't find any equivalent to the 'Subword Information' paper's P
in fastText.

But rather than modifying computeNgrams source, it should be enough to
just provide an extreme -minn or -maxn parameter.

A minn larger than 'maxn' should mean line https://github.com/
facebookresearch/fastText/blob/55ba6b505004539d3621d8eb5a10d5
259faa4bb2/src/dictionary.cc#L128 is never reached, so no word ever
decomposed into char n-grams, so the fastText code equivalent to classic
word-only word2vec. So running fastText with -maxn 0 seems sufficient to
disable char n-grams.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#16 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/ATll0HRFVKycnJQYRPBkhM8A35lmynUhks5qdmPKgaJpZM4JelKy
.

from fasttext.

Hi,

You are right, the parameter was removed from the final implementation. Given that the code is much more efficient not than then, this parameter was not critical. We may consider re-implementing it later if useful applications are found.

Best,

Piotr

from fasttext.

I think it should be done now it's not about how many codes are there it is
ABOUT the site security because if anyone else finds out you have a hole in
your site it will bring hackers to test their skills and that is nither
good for the server on the site so I suggest it should be fixed latest
this week
On Aug 8, 2016 10:23 AM, "Piotr Bojanowski" [email protected]
wrote:

Hi,

You are right, the parameter was removed from the final implementation.
Given that the code is much more efficient not than then, this parameter
was not critical. We may consider re-implementing it later if useful
applications are found.

Best,

Piotr

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#16 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/ATll0BOjrm2kV25ruk4ksSThsT6PB1OKks5qdvWVgaJpZM4JelKy
.

from fasttext.

Thanks a lot for the helpful responses, @gojomo and @piotr-bojanowski

from fasttext.