richtr/guesslanguage.js

A natural language detection library based on trigram statistical analysis for Node.js and the Web.

guesslanguage.js's Issues

Eample code not working.

In example HTML, if i remove the entire text and add something else or the same text it always return unknown... which makes me think that this plugin doesn't work at all?

document trigram building process

It would be nice if the trigram building process was documented. For example, the German data contains strings like didiescheincheichdenin which doesn't look like a trigram. Also, is it known what was used as input originally?

When trying to add it to angular it complains about CommonJS

https://angular.io/guide/build#configuring-commonjs-dependencies

By the way. Have used the library before and super satisfied with it to catch spam.

Best regards,
Sam

detection of zh_TW and zh_CN

This text, copied from http://www.gov.tw is detected as being zh.
駕駛執照替代役水質補(捐)助資訊出國疫苗地價稅牌照稅退休金技能檢定生育補助交通事故天災事變生活扶助健康檢查消費者身心障礙健康保險勞工保險教育補助
It would be nice if the detection was more specific (zh-TW).
The documentation claims to detect "zh": Chinese and "zh-TW" Chinese (Taiwan).

Burmese will never match

Burmese is listed as "Burmese" in SINGLETONS, but as "Myanmar" in unicodeBlockTests.

Fix: In unicodeBlockTests, instead of using "Myanmar" as the property, use "Burmese".

EDIT: Add fix.

What is the minimum amount of characters for guessLanguage to actually detect a language?

As I asked the dev on Twitter, I just wanted to know how many characters was needed for the program to detect a language (because song titles for example don’t seem to work).

_{Sent with GitHawk}

Source of language datasets

Where is the source text dataset for the Ngrams of those 100 languages? Would like to see if it is different from wooorm/franc#78 usage of UDHR, and if it is more accurate than them.

Strange exports of module in node

After install with NPM npm install guesslanguage
I have strange behaviour when importing.
Seems like the module exports a JS object so i have to duplicate and write guessLanguage.guessLanguage.detect

here an example in coffeescript

guessLanguage = require 'guesslanguage'
guessLanguage.guessLanguage.detect text, (language) ->
        console.log "language is : "+language

Change `richtr.github.com` to `richtr.github.io` at link demo README.md

At this line, can you correct the link to https://richtr.github.io/guessLanguage.js/?

Because at current time, it is a 404 page.
https://github.com/richtr/guessLanguage.js/blob/master/README.md?plain=1#L8

Licensing and the Quail project

We are discussing using guessLanguage in the Quail accessibility project here:

quailjs/quail#58

One issue is that of license incompatibility, since the project is MIT. We are really interested in using your existing trigram database, as there were some edge cases we have to capture with our own language-guessing project. Would you be amenable to us including the trigram database in the project, or discuss other options?

Wrong Classification for an English Text

I tried this text in the demo page
aim at your goals, how many words do you want I can't understand?????

And the result is:
Detected language of provided text is Afrikaans [af].

Lack a list of supported languages in the README file

It will be useful to have such a list at a glance.
Thank you very much for the good job :)

Norsk will never match

Norsk (Norwegian) is listed in the trigram database as "nb" (Bokmål), but as "no" (Both Bokmål and Nynorsk) in the guessLanguage.js file.

A fix would be to rename the property in the trigram database ("nb" to "no"), but that might result in incorrect results, although the difference between Bokmål and Nynorsk is not that big for the trigram. A better fix would probably be (without changing the database) to classify as "nb" instead of "no".