def __init__(self, triple_path):
self.entity_set = set() # 实体节点集合
self.nlp = NLP()
# 连接neo4j数据库
self.graph = Graph(host='localhost',
http_port=7474,
user='******',
password='******')
f_in = open(triple_path, 'r')
triple_str = f_in.read() # 读取整个Json
self.triple = json.loads(triple_str)
sys.path.append("..") # 先跳出当前目录
from core.nlp import NLP
from core.extractor import Extractor
if __name__ == '__main__':
input_path = '../../data/input_text.txt' # 输入的文本文件
output_path = '../../data/knowledge_triple.json' # 输出的处理结果Json文件
if os.path.isfile(output_path):
os.remove(output_path)
# os.mkdir(output_path)
print('Start extracting...')
# 实例化NLP(分词，词性标注，命名实体识别，依存句法分析)
nlp = NLP()
num = 1 # 知识三元组
with open(input_path, 'r', encoding='utf-8') as f_in:
# 分句，获得句子列表
origin_sentences = re.split('[。？！；]|\n', f_in.read())
# 遍历每一篇文档中的句子
for origin_sentence in origin_sentences:
# 原始句子长度小于6，跳过
if (len(origin_sentence) < 6):
continue
print('*****')
# print(origin_sentence)
# 分词处理
lemmas = nlp.segment(origin_sentence)
class WriteToNeo4j:
"""将Json类型的知识三元组导入Neo4j数据库"""
def __init__(self, triple_path):
self.entity_set = set() # 实体节点集合
self.nlp = NLP()
# 连接neo4j数据库
self.graph = Graph(host='localhost',
http_port=7474,
user='******',
password='******')
f_in = open(triple_path, 'r')
triple_str = f_in.read() # 读取整个Json
self.triple = json.loads(triple_str)
def write_litigant(self, litigants, relation):
"""处理当事人信息(原告和原告)
Args:
litigant: list，当事人信息
"""
for litigant in litigants:
node_litigant = Node(self.get_label(litigant['名字']),
name=litigant['名字'],
id=litigant['编号'])
self.graph.create(node_litigant)
self.entity_set.add(litigant['名字'])
node_root = self.graph.find_one('判决书',
property_key='name',
property_value='判决书001')
entity_relation = Relationship(node_root,
relation,
node_litigant,
label='relation')
self.graph.create(entity_relation)
for item in litigant:
if item != '名字' and item != '编号':
node_repr = Node(self.get_label(litigant[item]),
name=litigant[item]) # 负责人，委托代理人
self.graph.create(node_repr)
self.entity_set.add(litigant[item])
entity_relation = Relationship(node_litigant,
item,
node_repr,
label='关系')
self.graph.create(entity_relation)
def get_label(self, word):
"""根据单词获得标签
Args:
word: str，单词
Returns:
label: str，类型标签
"""
label = ''
postag = self.nlp.get_postag(word)
if postag == 'nh':
label = '人'
elif postag == 'ni':
label = '组织'
elif postag == 'ns':
label = '地点'
else:
label = '其他'
return label
def write(self):
"""写入图数据库"""
# 根节点
# 一篇判决书具有"文书编号"，"文书标题"，"按键编号"，"文书类型"，"案件编号"几个属性
node_root = Node('判决书',
name='判决书001',
id=self.triple['文书编号'],
title=self.triple['文书标题'],
type=self.triple['文书类型'],
case=self.triple['案件编号'])
self.graph.create(node_root)
self.entity_set.add('判决书001')
node_court = Node('组织', name=self.triple['受理法院'])
self.graph.create(node_court)
self.entity_set.add(self.triple['受理法院'])
entity_rerlation = Relationship(node_root,
'受理法院',
node_court,
label='关系')
self.graph.create(entity_rerlation)
# 遍历原告，被告
plaintiffs = self.triple['原告']
self.write_litigant(plaintiffs, '原告')
defendants = self.triple['被告']
self.write_litigant(defendants, '被告')
facts = self.triple['案情事实']
for fact in facts:
tri = fact['知识']
entity1 = tri[0]
relation = tri[1]
entity2 = tri[2]
node_list = []
node1 = Node(self.get_label(entity1), name=entity1)
if entity1 not in self.entity_set:
self.graph.create(node1)
node_list.append(node1)
self.entity_set.add(entity1)
else:
node_list.append(
self.graph.find_one(self.get_label(entity1),
property_key='name',
property_value=entity1))
node2 = Node(self.get_label(entity2), name=entity2)
if entity2 not in self.entity_set:
self.graph.create(node2)
node_list.append(node2)
self.entity_set.add(entity2)
else:
node_list.append(
self.graph.find_one(self.get_label(entity2),
property_key='name',
property_value=entity2))
entity_relation = Relationship(node_list[0],
relation,
node_list[1],
label='关系')
self.graph.create(entity_relation)
# filter_str('acbdDEF哈哈哈🍕')
if __name__ == '__main__':
# input_path = '../../data/input_text.txt' # 输入的文本文件
input_path = '../../data/第三次减半后内容.txt' # 输入的文本文件
output_path = '../../data/knowledge_triple6.json' # 输出的处理结果Json文件
if os.path.isfile(output_path): ###删除指定路径下的文件
os.remove(output_path)
# os.mkdir(output_path)
print('Start extracting...')
# 实例化NLP(分词，词性标注，命名实体识别，依存句法分析)
nlp = NLP()
num = 1 # 知识三元组
with open(input_path, 'r', encoding='utf-8') as f_in:
# 分句，获得句子列表
origin_sentences = re.split('[。？！；]|\n', f_in.read())
# 遍历每一篇文档中的句子
for origin_sentence in origin_sentences:
# 原始句子长度小于2，跳过
if (len(origin_sentence) < 2):
continue
# print('原始句子:',origin_sentence)
origin_sentence = filter_str(origin_sentence)
# print('*****')
# print('处理句子:',origin_sentence)
# print('type:',type(origin_sentence))
# 分词处理 jieba分词工具