wswu/yawipa

A comprehensive and extensible Wiktionary parsing framework.

Yawipa

A comprehensive and extensible Wiktionary parsing framework. If we don't have a parser for your edition of Wiktionary, please help contribute one!

Usage

julia yawipa.jl --dump DUMP --out OUT --log LOG --parsers all

where DUMP is the Wiktionary XML dump (e.g. this one for the English Wiktionary).

The argument to --parsers is a comma-separated list of edition-specific parsing functions defined in the respective parsers/{lang}.jl, or "all" to use all available parsing functions.

For advanced filtering, you can specify --skip SKIP to skip page titles matching the specified regex.

Data

If you don't want to run the parser and just want the data, look here! These are TSV files that are easy to use in your own applications. They will be updated monthly with the latest Wiktionary dumps.

Writing your own parser

It is simple to write a new parser for a Wiktionary edition. The parser for the Italian Wiktionary at src/parsers/it.jl is a good example of a barebones parser. There are a few things to keep in mind:

• Each parser should be in its own module. The convention (which we might change in the future) is that the module name is the camelcased language code in Wiktionary (e.g. module En for the English edition). This module will be programatically imported based on the --edition argument.
• The parser must be called {lang}Parser, replacing {lang} with this language code.
• The parser must have a lang_from_heading function which identifies the language from the heading of the entry. If the heading is not a language code, this function should return nothing. Some editions will need a table lookup to convert the language name to a language code.
• Then you can define a parsing function for whatever you want to parse! This function must take four arguments (language, page title, section heading, section text). It should return a list of parsed information to be written to the specified output file. This function should be added to the parsing_functions dictionary. Typically you will be parsing Wiktionary templates (e.g. {{t|yue|字典|tr=zi6 din2}}), and there are a variety of functions in src/template.jl that facilitate this parsing.

Citation

If you found this software useful, please cite

@inproceedings{wu-yarowsky-2020-yawipa,
    title = "Computational Etymology and Word Emergence",
    author = "Wu, Winston and Yarowsky, David",
    booktitle = "Proceedings of The 12th Language Resources and Evaluation Conference",
    month = may,
    year = "2020",
    address = "Marseille, France",
    publisher = "European Language Resources Association",
    url = "https://www.aclweb.org/anthology/2020.lrec-1.397",
}

If you use the extracted morphological data or translations from etymology glosses, please also cite

@inproceedings{wu-yarowsky-2020-wiktionary,
    title = "{W}iktionary Normalization of Translations and Morphological Information",
    author = "Wu, Winston and Yarowsky, David",
    booktitle = "Proceedings of the 28th International Conference on Computational Linguistics",
    month = dec,
    year = "2020",
    address = "Barcelona, Spain (Online)",
    publisher = "International Committee on Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2020.coling-main.413",
}