基于neo4j,django,pytorch,py2neo的兽药知识图谱及问答（1）

系统参考的项目地址：https://github.com/qq547276542/Agriculture_KnowledgeGraph

django项目的搭建

使用pycharm创建django项目。项目结构：

• kg
  • appdemo
    + models // 模型层，用于和neo4j交互，实现查询等核心功能
    + views// 用于写页面的逻辑(View)
  • statics// 静态资源
  • templates// html页面

实体查询

得到前端输入的实体，预处理，match语句进行匹配，返回和该实体相关的节点信息等。

def search_entity(request):
	ctx = {
    }
	#根据传入的实体名称搜索出关系
	if(request.GET):
		entity = request.GET['user_text']
		entity = entity.strip()
		entity = entity.lower()
		entity = ''.join(entity.split())

		actor_name_dict, movie_name_dict = name_dict[0], name_dict[1]

		if entity in actor_name_dict.keys():
			entity = actor_name_dict.get(entity)
		if entity in movie_name_dict.keys():
			entity = movie_name_dict.get(entity)

		entityRelation = neo4jconn.get_entity_info(entity)
		if len(entityRelation) == 0:
			#若数据库中无法找到该实体，则返回数据库中无该实体
			ctx= {
    'title' : '<h2>知识库中暂未添加该实体</h1>'}
			return render(request,'entity_search.html',{
    'ctx':json.dumps(ctx,ensure_ascii=False)})
		else:
			return render(request,'entity_search.html',{
    'entityRelation':json.dumps(entityRelation,ensure_ascii=False)})
	#需要进行类型转换
	return render(request, 'entity_search.html', {
    'ctx':ctx})

match语句，写在model中的。使用py2neo连接neo4j图数据库。返回json数据，echarts解析并绘制节点关系图。

	def connectNeo4j(self):
		self.graph = Graph("http://127.0.0.1:7474", username="neo4j", password="root")
		self.matcher = NodeMatcher(self.graph)

	# 一.实体查询
	def get_entity_info(self, name) -> list:
		'''
        查找该entity所有的直接关系
        :param name:
        :return:
        '''

		data = self.graph.run(
			"match (source)-[rel]-(target)  where source.name = $name " +
			"return rel ", name=name).data()

		json_list = []
		for an in data:
			result = {
    }
			rel = an['rel']
			relation_type = list(rel.types())[0]
			start_name = rel.start_node['name']
			end_name = rel.end_node['name']
			result["source"] = start_name
			result['rel_type'] = relation_type
			result['target'] = end_name
			json_list.append(result)

		return json_list

关系查询

和实体查询类似，关系查询仅查询直接1度关系。考虑不同的情况，仅包含头实体1，仅包含尾实体2，实体1+关系，关系+实体2，实体1+实体2，实体1+关系+实体2

# 6.关系查询：实体1+关系+实体2(实体-关系->实体)
	def findEntityRelation(self,entity1,relation,entity2):
		answer = self.graph.run(
			"match (source)-[rel:" + relation + "]->(target)  where source.name= $name1 and target.name = $name2 " +
			"return rel ", name1=entity1, name2=entity2).data()

		answer_list = []
		for an in answer:
			result = {
    }
			rel = an['rel']
			relation_type = list(rel.types())[0]
			start_name = rel.start_node['name']
			end_name = rel.end_node['name']
			result["source"] = {
    'name': start_name}
			result['type'] = relation_type
			result['target'] = {
    'name': end_name}
			answer_list.append(result)

		return answer_list

问答