python-mecab-ner

mecab_ner

Table of Contents

• Features
• Installation
• Quick start
• Simple Example
• Practical Example

Features

• Python library to get NER using Mecab
• Infer NER from Data dictionary

Installation

Download using pip via pypi.

$ pip install 'python-mecab-ner' --upgrade

(Mac/homebrew users may need to use pip3)

Simple Example

from python_mecab_ner.mecab_ner import MecabNer

mecab_ner = MecabNer()
test_sentence = "아이유의 금요일에 만나요를 들으면서 신촌 딸기를 먹을래"

mecab_ner.parse(test_sentence)
# [('아이유', NerFeature(word='아이유', pos='ner', category=Category(large='ner_example_music_singer', small='가수'))), ('의', NerFeature(word='의', pos='JKG', category=None)), ('금요일에 만나요', NerFeature(word='금요일에 만나요', pos='ner', category=Category(large='ner_example_music_song', small='노래'))), ('를', NerFeature(word='를', pos='JKO', category=None)), ('듣', NerFeature(word='듣', pos='VV+EC', category=None)), ('으면서', NerFeature(word='으면서', pos='VV+EC', category=None)), ('신촌 딸기', NerFeature(word='신촌 딸기', pos='ner', category=Category(large='ner_example_fruit', small='과일'))), ('를', NerFeature(word='를', pos='JKO', category=None)), ('먹', NerFeature(word='먹', pos='VV', category=None)), ('을래', NerFeature(word='을래', pos='EC', category=None))]

mecab_ner.morphs(test_sentence)
# ['아이유', '의', '금요일에 만나요', '를', '듣', '으면서', '신촌 딸기', '를', '먹', '을래']

mecab_ner.ners(test_sentence)
# [('아이유', 'ner_example_music_singer', '가수'), ('금요일에 만나요', 'ner_example_music_song', '노래'), ('신촌 딸기', 'ner_example_fruit', '과일')]

Practical Example

• Set data path for your work
• File name will become entity set category large

Data Directory

# Directory
root/
    word_dir.py
    data/
        programming.txt
        public.txt

Data File

• Data file must be txt format and first line should start your small category
• First line will become entity set category small

- data/programming.txt
#인공지능
파이썬
딥러닝
머신 러닝
자연어 처리
버트
#백엔드
로그
http통신

- data/place.txt
#hospital
병원

Results

• on example code 1 ('자연어 로그', 'programming', '백엔드') is infered ner from vocab 로그 in data file.
• on example code 2 ('서울대병원', 'place', 'hospital') is infered ner from vocab 병원 in data file.
• if you don't want, set MecabNer(infer=False)

# example code 1
from python_mecab_ner.mecab_ner import MecabNer

mecab_ner = MecabNer(ner_path="./data")
test_sentence = "자연어 처리를 위해 인공지능을 위한 파이썬을 공부하여 자연어와 관련된 일을 하고 있습니다. http 요청시 자연어 로그를 쌓는 것이 중요합니다."

mecab_ner.parse(test_sentence)
# [('자연어 처리', NerFeature(word='자연어 처리', pos='ner', category=Category(large='programming', small='인공지능'))), ('를', NerFeature(word='를', pos='JKO', category=None)), ('위하', NerFeature(word='위하', pos='VV+EC', category=None)), ('아', NerFeature(word='아', pos='VV+EC', category=None)), ('인공', NerFeature(word='인공', pos='NNP', category=None)), ('지능', NerFeature(word='지능', pos='NNP', category=None)), ('을', NerFeature(word='을', pos='JKO', category=None)), ('위하', NerFeature(word='위하', pos='VV+ETM', category=None)), ('ᆫ', NerFeature(word='ᆫ', pos='VV+ETM', category=None)), ('파이썬', NerFeature(word='파이썬', pos='ner', category=Category(large='programming', small='인공지능'))), ('을', NerFeature(word='을', pos='JKO', category=None)), ('공부', NerFeature(word='공부', pos='NNG', category=None)), ('하', NerFeature(word='하', pos='XSV', category=None)), ('여', NerFeature(word='여', pos='EC', category=None)), ('자연', NerFeature(word='자연', pos='NNG', category=None)), ('어', NerFeature(word='어', pos='NNG', category=None)), ('와', NerFeature(word='와', pos='JC', category=None)), ('관련', NerFeature(word='관련', pos='NNG', category=None)), ('되', NerFeature(word='되', pos='XSV+ETM', category=None)), ('ᆫ', NerFeature(word='ᆫ', pos='XSV+ETM', category=None)), ('일', NerFeature(word='일', pos='NNG', category=None)), ('을', NerFeature(word='을', pos='JKO', category=None)), ('하', NerFeature(word='하', pos='VV', category=None)), ('고', NerFeature(word='고', pos='EC', category=None)), ('있', NerFeature(word='있', pos='VX', category=None)), ('습니다', NerFeature(word='습니다', pos='EF', category=None)), ('.', NerFeature(word='.', pos='SF', category=None)), ('http', NerFeature(word='http', pos='SL', category=None)), ('요청', NerFeature(word='요청', pos='NNG', category=None)), ('시', NerFeature(word='시', pos='NNB', category=None)), ('자연어 로그', NerFeature(word='자연어 로그', pos='ner', category=Category(large='programming', small='백엔드'))), ('를', NerFeature(word='를', pos='JKO', category=None)), ('쌓', NerFeature(word='쌓', pos='VV', category=None)), ('는', NerFeature(word='는', pos='ETM', category=None)), ('것', NerFeature(word='것', pos='NNB', category=None)), ('이', NerFeature(word='이', pos='JKS', category=None)), ('중요', NerFeature(word='중요', pos='NNG', category=None)), ('하', NerFeature(word='하', pos='XSV+EF', category=None)), ('ᄇ니다', NerFeature(word='ᄇ니다', pos='XSV+EF', category=None)), ('.', NerFeature(word='.', pos='SF', category=None))]

mecab_ner.morphs(test_sentence)
# ['자연어 처리', '를', '위하', '아', '인공', '지능', '을', '위하', 'ᆫ', '파이썬', '을', '공부', '하', '여', '자연', '어', '와', '관련', '되', 'ᆫ', '일', '을', '하', '고', '있', '습니다', '.', 'http', '요청', '시', '자연어 로그', '를', '쌓', '는', '것', '이', '중요', '하', 'ᄇ니다', '.']

mecab_ner.ners(test_sentence)
# [('자연어 처리', 'programming', '인공지능'), ('파이썬', 'programming', '인공지능'), ('자연어 로그', 'programming', '백엔드')]

test_sentence2 = "나는 서울대병원에 갈려고 합니다."

mecab_ner.parse(test_sentence2)
# [('나', NerFeature(word='나', pos='NP', category=None)), ('는', NerFeature(word='는', pos='JX', category=None)), ('서울대병원', NerFeature(word='서울대병원', pos='ner', category=Category(large='place', small='hospital'))), ('에', NerFeature(word='에', pos='JKB', category=None)), ('가', NerFeature(word='가', pos='VV+EC', category=None)), ('ᆯ려고', NerFeature(word='ᆯ려고', pos='VV+EC', category=None)), ('하', NerFeature(word='하', pos='VX+EF', category=None)), ('ᄇ니다', NerFeature(word='ᄇ니다', pos='VX+EF', category=None)), ('.', NerFeature(word='.', pos='SF', category=None))]

mecab_ner.morphs(test_sentence2)
# ['나', '는', '서울대병원', '에', '가', 'ᆯ려고', '하', 'ᄇ니다', '.']

mecab_ner.ners(test_sentence2)
# [('서울대병원', 'place', 'hospital')]