Help

Lists

Revision as of 09:53, 20 May 2018 by Yug (talk | contribs) (Created page with "Wordlists are at the core of language protection. LinguaLibre has a dedicated namespace to store and document lists. For how to on how to process lists, see Help:Main. ==...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Wordlists are at the core of language protection. LinguaLibre has a dedicated namespace to store and document lists. For how to on how to process lists, see Help:Main.

Guidelines

  1. Find or create an open source wordlist
  2. Format it so each line has one single item, be it a word or phrase. Prefixes - or * are valid.
  3. Create a wikipage with name List:{iso-639-3}/{topic's_title} (ex: List:Fra/Fruits.
  4. * Title in target language or English.
Template

List:eng/legendary_creatures

- dragon
- unicorn
- dahu
- phoenix
- centaur
- pegasus
=== Source ===
* {Source citation}. {License}

Collaboration policies

  • Lists can be expanded, edited, replaced, moved.
  • If on same language, same topic : add the author's name.

Licenses

Public domain, MIT, GPL, open Creative commons (cc-by, cc-by-sa) are acceptable.

Access wordlists

Finding wordlists

As of 2018, a large amount of open source wordlists are available online for most major languages.

LinguaLibre provides Sparkles queries to generate wordlists based on the nearly 300 wikipedias.

Hand editing wordlists

Wordlists copyrights are rather lose. Exact copy of copyrighted dictionaries indexes or wordlists is forbidden. But a new wordlist made from several sources by picking up individuals words, which are not individually copyrighted, is itself your own work.

Datamining

Taking as input a list of text files in the target language (aka "written corpus"), it is possible, for many languages, to create wordlists via terminal and bash commands. Expect half a day of work. Depending on the language, this can present different challenges : East Asian language don't have spaces marking clear word separation, other languages have rich agglomerative behaviors so meaningful roots get diluted. For these, more NLP or hand editin is required.