https://github.com/kshramt/company-ngram

https://github.com/kshramt/company-ngram

completion-candidates n-grams natural-language-processing

Last synced: 1 day ago
JSON representation

• Host: GitHub
• URL: https://github.com/kshramt/company-ngram
• Owner: kshramt
• Created: 2016-03-04T04:55:40.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2017-03-20T12:53:25.000Z (over 7 years ago)
• Last Synced: 2024-08-02T05:11:59.922Z (3 months ago)
• Topics: completion-candidates, n-grams, natural-language-processing
• Language: Python
• Size: 105 KB
• Stars: 30
• Watchers: 4
• Forks: 3
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# company-ngram

A company backend for N-gram based completion.

![](screenshot.jpg)

This backend produces completion candidates that are fuzzily matching to N-gram data.

The N-gram data is automatically constructed from `*.txt` files placed directly under `company-ngram-data-dir` directory.

If you set `company-ngram-n` to `4`, three words before the cursor are used to produce completion candidates.

To mitigate the data sparsity problem, this backend uses a fuzzy-matching strategy.

Given the following sentence, `Dear Dr. Aki, `, this backend produces completion candidates that match at least one of following prefixes,

```

Dear Dr. Aki,

*    Dr. Aki,

Dear *   Aki,

*    *   Aki,

Dear Dr. *

*    Dr. *

Dear *   *

```

where `*` matches an arbitrary word.

Hence, even if your `*.txt` does not contain the word `Aki`, you still have chance to get completion candidates.

## Configurations

```elisp

; ~/.emacs.d/init.el

(with-eval-after-load 'company-ngram

  ; ~/data/ngram/*.txt are used as data

  (setq company-ngram-data-dir "~/data/ngram")

  ; company-ngram supports python 3 or newer

  (setq company-ngram-python "python3")

  (company-ngram-init)

  (cons 'company-ngram-backend company-backends)

  ; or use `M-x turn-on-company-ngram' and

  ; `M-x turn-off-company-ngram' on individual buffers

  ;

  ; save the cache of candidates

  (run-with-idle-timer 7200 t

                       (lambda ()

                         (company-ngram-command "save_cache")

                         ))

  )

(require 'company-ngram nil t)

```

[RFC](http://www.rfc-editor.org/rfc-index.html) provides handy text files for a quick trial.

```bash

wget --directory-prefix ~/data/ngram    https://www.rfc-editor.org/rfc/rfc{5661,6716,4949}.txt

```

## License

[The GNU General Public License version 3](http://www.gnu.org/licenses/).