PaddlePaddle/PaddleNLP

[Question]: UIE模型做信息抽取时如何设置嵌套关系的schema

Opened this issue · 1 comments

请提出你的问题

我在用UIE-base模型做信息抽取,现在信息标注到抽取都跑通了,准确率也不错。但是,现在遇到个问题,我想设置嵌套关系的schema,但是好像UIE不支持,请问应该怎么调整?我想用“储层”作为“油田区块”的子实体,结果报错如下:

schema ={'油田区块': {'储层': ['地面原油密度', '地面原油粘度', '含硫量', '含蜡量']}}
ie.set_schema(schema)
Traceback (most recent call last):
File "", line 1, in
File "/usr/local/lib/python3.10/dist-packages/paddlenlp/taskflow/taskflow.py", line 862, in set_schema
self.task_instance.set_schema(schema)
File "/usr/local/lib/python3.10/dist-packages/paddlenlp/taskflow/information_extraction.py", line 549, in set_schema
self._schema_tree = self._build_tree(schema)
File "/usr/local/lib/python3.10/dist-packages/paddlenlp/taskflow/information_extraction.py", line 1307, in _build_tree
raise TypeError(
TypeError: Invalid schema, value for each key:value pairs should be list or stringbut <class 'dict'> received

UIE不支持嵌套实体识别,需要做一些后处理策略,目前没有相关的开发计划,欢迎开发者贡献。

#4503