langchain-jieba

An integration package connecting Jieba and LangChain

Jieba link extractor

Description

langchain_community.graph_vectorstores support storing texts and their relationships in a vector database in the form of a knowledge graph.But the recommended extractor such as KeybertLinkExtractor works better in an English environment,while it performs poorly in a Chinese environment. JiebaLinkExtractor extract keywords from texts depend on jieba.analyse lib, which more suitable in a Chinese environment.

Installation

%pip install langchain_jieba

Usage

extract from single text

from langchain_jieba import JiebaLinkExtractor

extractor = JiebaLinkExtractor()
results = extractor.extract_one("小明硕士毕业于中国科学院计算所，后在日本京都大学深造")
print(f"results={results}")

results={Link(kind='kw', direction='bidir', tag='日本京都大学'), Link(kind='kw', direction='bidir', tag='计算所'), Link(kind='kw', direction='bidir', tag='小明')}

extract from multiple texts

from langchain_jieba import JiebaLinkExtractor

CONTENT1 = '小明硕士毕业于中国科学院计算所，后在日本京都大学深造'
CONTENT2 = '我来到北京清华大学'
extractor = JiebaLinkExtractor()
results = list(extractor.extract_many([CONTENT1, CONTENT2]))
print(f"results[0]={results[0]}")
print(f"results[1]={results[1]}")

results[0]={Link(kind='kw', direction='bidir', tag='日本京都大学'), Link(kind='kw', direction='bidir', tag='计算所'), Link(kind='kw', direction='bidir', tag='小明')} results[1]={Link(kind='kw', direction='bidir', tag='清华大学'), Link(kind='kw', direction='bidir', tag='来到'), Link(kind='kw', direction='bidir', tag='北京')}