solver-ai / textrankr

TextRank for Korean.

textrankr

Reorder sentences using the TextRank algorithm.

• Mostly designed for Korean, but not limited to.
• Check out lexrankr, which is an another awesome summarizer!
• Not available for Python 2 anymore (if necessary, use version 0.3).

Installation

pip install textrankr

Tokenizers

Tokenizers are not included. You have to implement it by yourself.

Example:

from typing import List

class MyTokenizer:
    def __call__(self, text: str) -> List[str]:
        tokens: List[str] = text.split()
        return tokens

한국어의 경우 KoNLPy를 사용하는 방법이 있습니다. 아래 예시처럼 phrases를 쓰게되면 엄밀히는 토크나이저가 아니지만 이게 더 좋은 결과를 주는것 같습니다.

from typing import List
from konlpy.tag import Okt

class OktTokenizer:
    okt: Okt = Okt()

    def __call__(self, text: str) -> List[str]:
        tokens: List[str] = self.okt.phrases(text)
        return tokens

Usage

from typing import List
from textrankr import TextRank

mytokenizer: MyTokenizer = MyTokenizer()
textrank: TextRank = TextRank(mytokenizer)

k: int = 3  # num sentences in the resulting summary

summarized: str = textrank.summarize(your_text_here, k)
print(summarized)  # gives you some text

# if verbose=False, it returns a list
summaries: List[str] = textrank.summarize(your_text_here, k, verbose=False)
for summary in summaries:
    print(summary)

Test

Use docker.

docker build -t textrankr -f Dockerfile .
docker run textrankr