Skip to content

Latest commit

 

History

History
108 lines (88 loc) · 3.31 KB

README.md

File metadata and controls

108 lines (88 loc) · 3.31 KB
Raw

Introduction

Warning: This software is currently alpha. Don't use the data blindly.

Kanji usage in Japan is regulated by the Jōyō Kanji table. The latest, 2010 edition of the table is provided by the Ministry of Education in PDF format: http://kokugo.bunka.go.jp/kokugo_nihongo/joho/kijun/naikaku/pdf/joyokanjihyo_20101130.pdf

The original table is quite messy and hard to use in computer programs. This project includes code to extract the data and convert it into popular formats: TSV, JSON, SQL and HTML. To minimize human error, the data is parsed automatically as much as possible. The results will be tested to ensure consistency.

Most users won't have to run the scripts to extract the data; you can just download the data directly from the output directory.

Roadmap/TODO

Completed

  • On- and kun-readings

    • Romaji converter.
    • Distinguish uncommon readings (marked in the table as indented/1字下げ).

  • Example words

    • Use examples to delimit okurigana in kun-readings
      • Including inflected examples, and "double okurigana" (like 成り立ち)
    • Handle POS markers :〔副〕,〔接〕, '……',
    • Treat glossed variations as different, special readings
    • Handle examples with explicative text and 「」

  • Old kanji

    • Handle 弁:[辨, 瓣, 辯].
    • Handle 亀/龜 as Unicode.

  • Variant forms

    • Encode accepted variants (許容字体) as Unicode variation sequences.
    • Convert little-used codepoints to popular use alternatives (通用字体: 塡 剝 頰 → 填 剥 頬).
    • Convert 叱 U+53F1 into common alternate (異体字) 𠮟 U+20B9F.
    • Reference images

  • Notes (参考)

    • Distinguish kanji-scoped notes from reading-scoped ones.
    • Save full note as text
      • Handle notes spanning multiple lines
    • Extract data from notes:
      • Alternate orthographies.
      • Compounds (test against appendix).
      • Prefecture names.
      • Examples marked as literary (文語).

  • Output formats

    • TSV

  • Tests

    • doctests for functions
    • old_kanji: against wikipedia, old dataset
    • readings: against kanjidic
    • examples: against JMdict (edict)

Yet to be done

  • Notes (参考)

    • Parse and structure the data from notes
      • Alternative orthographies (同訓異字).
      • Pointers to reference section in text.
        • Extract example images.
      • Exceptional readings..
        • Unbounded lists (with a "などは").
          • Complement with all available examples from edict.
        • Alternatives ("とも").
      • One-off types of notes.

  • Tests

    • notes: write at least one test for each type of parsed note

  • Parse appendix

  • Output types:

    • SQL
    • JSON
    • HTML table

  • Document:

How to recreate the files

 pip3 install romkan
 pip3 install ostruct
 pip3 install regex # newer version of 're'
 git clone https://github.com/leoboiko/joyodb.git
 cd joyodb
 make # (needs Internet)
 bin/convert_joyodb

Output will be in output/ directory.

How to test

apt-get install rsync python3-lxml python3-bs4 mecab unidic-mecab
pip3 install mecab-python3
make test # (needs Internet)