ja-entity-parser 0.2.0

Japanese entity parser library for company/corporate name normalization and extraction.

ja-entity-parser

日本語 / English

Overview

ja-entity-parser is a Python library for normalization and extraction of Japanese entities such as company names, corporate types, personal names, and addresses.
It combines SudachiPy morphological analysis with custom normalization rules (old/new kanji conversion, bracket/punctuation/control character unification, NFKC, and user dictionary replacements) to accurately extract brand names and legal forms (e.g., 株式会社, 合同会社).

Features

• Japanese text normalization: Old/new kanji conversion, bracket/punctuation/control character unification, NFKC, custom dictionary replacements
• Company/corporate type extraction: Uses SudachiPy and part-of-speech info
• Katakana reading output: Builds brand_kana by concatenating token reading forms
• User dictionary support: Extendable for industry-specific terms
• Testing & extensibility: Comes with pytest-based unit tests

Installation

pip install ja-entity-parser

Usage

1. Extract company/corporate info (normalize + parse)

from ja_entityparser import corporate_parser

result = corporate_parser("トヨタ自動車株式会社")
print(result)
# {'input': 'トヨタ自動車株式会社', 'legal_form': '株式会社', 'brand_name': 'トヨタ自動車', 'brand_kana': 'トヨタジドウシャ'}

2. Normalization only

from ja_entityparser.normalizer import normalize

text = "〔トヨタ〕株式会社"
print(normalize(text))
# (トヨタ)株式会社

3. Parsing only (pass normalized text)

from ja_entityparser.parser import parse

result = parse("トヨタ自動車株式会社")
print(result)
# {'legal_form': '株式会社', 'brand_name': 'トヨタ自動車', 'brand_kana': 'トヨタジドウシャ'}

API

• corporate_parser(text: str) -> dict
  • Normalize and parse input, returns {'input': ..., 'legal_form': ..., 'brand_name': ..., 'brand_kana': ...} (brand_kana when available)
• normalize(text: str) -> str
  • Normalize Japanese text
• parse(text: str) -> dict
  • Morphological analysis and extraction of brand name/legal form (and brand_kana when available)

License

Apache License 2.0