This is a script to convert the output file of doccano
to a format that is easy to handle with sklearn-crfsuite
.
python doccano.py --input_path <path to doccano exported jsonl file> --output_path <path to output file>
The input file is a jsonl file exported from doccano
.
{"text": "東京都渋谷区渋谷 2丁目2−8 渋谷マークシティ", "labels": [[0, 9, "LOC"]]}
{"text": "東京都渋谷区神南 1丁目1−1", "labels": [[0, 7, "LOC"]]}
...
The output file is a json file of the following format:
[
[
{"word": "東京都", "label": "B-LOC", "pos_tag": "名詞", "pos_tag[:2]": "名詞,固有名詞", "pos_tag_all": "名詞,固有名詞,地域,一般,*,*,東京都,トウキョウト,トーキョート", "BOS": true, "EOS": false},
{"word": "渋谷区", "label": "I-LOC", "pos_tag": "名詞", "pos_tag[:2]": "名詞,固有名詞", "pos_tag_all": "名詞,固有名詞,地域,一般,*,*,渋谷区,シブヤク,シブヤク", "BOS": false, "EOS": false},
...
],
...,
]
This program is mainly based on the following repository. https://github.com/ToshihikoSakai/jsontoconll
All mistakes in this script are mine.