基于深度学习的食品安全事件实体自动抽取模型研究

Research of Automatic Extraction of Entities from of Food Safety Event based on Deep Learning Model

实体在非结构化文本中不仅与词汇具有密切的关系,而且是构成短语的关键部分,特别是实体自身具有丰富的语义性,能够为后续语义知识的深度挖掘奠定基础。为了更好地从食品安全事件文本中挖掘出更加有价值和意义的知识,结合LSTM-CRF模型,笔者提出了食品安全事件实体抽取的基本流程,并构建了相应的食品安全事件实体抽取模型。在选取领域食品安全事件文本上,构建的食品安全事件实体抽取模型的调和平均值达到了相对可以接受的程度。基于深度学习的食品安全事件实体自动抽取模型不仅为实体的抽取提供了策略,而且在一定程度上验证了深度学习性能的整体状况。

Entities in unstructured texts are not only closely related to vocabulary,but also a key part of phrases.Especially,entities themselves have rich semantics,which can lay a foundation for further mining of semantic knowledge.In order to mine more valuable and meaningful knowledge from the text of food safety incidents,combined with LSTM-CRF model,the author proposed the basic process of entity extraction of food safety incidents,and constructed the corresponding entity extraction model of food safety incidents.In the text of food safety incidents in the field of selection,the harmonic average of the entity extraction model of food safety incidents has reached a relatively acceptable level.The food safety event entity automatic extraction model based on in-depth learning not only provides a strategy for entity extraction,but also verifies the overall situation of in-depth learning performance to a certain extent.