English

中国农机化学报

中国农机化学报 ›› 2023, Vol. 44 ›› Issue (2): 99-105.DOI: 10.13733/j.jcam.issn.2095-5553.2023.02.014

• 农业信息化工程 • 上一篇    下一篇

基于改进Bi-LSTM-CRF的农业问答系统研究#br#
#br#

白皓然1,孙伟浩1,金宁2,马皓冉1   

  1. 1. 青岛农业大学机电工程学院,山东青岛,266109; 2. 沈阳建筑大学研究生院,沈阳市,110168
  • 出版日期:2023-02-15 发布日期:2023-02-28
  • 基金资助:
    广东省重点领域研发计划(2018B020241003);山东省农机装备研发创新计划(2018YZ002);国家级大学生创新创业训练计划(S202010435013、201810435050X)

Research on agricultural question answering system based on improved Bi-LSTM-CRF

Bai Haoran, Sun Weihao, Jin Ning, Ma Haoran.   

  • Online:2023-02-15 Published:2023-02-28

摘要: 针对农业领域问答系统面临的实体识别困难的问题,提出一种基于改进Bi-LSTM-CRF的实体识别方法。首先通过BERT预训练模型的预处理,生成基于上下文信息的词向量,然后将训练出的词向量输入Bi-LSTM-CRF做进一步的训练处理,最后,利用Python的Django框架设计农业领域的实体识别、实体查询、农知问答等子系统。经过试验对比,所提出的改进的Bi-LSTM-CRF在农业信息领域具有更好的实体识别能力,在农业信息语料库上的精确率、召回率和F1值分别为93.23%、91.08%和92.16%。实现农业领域实体识别和农业信息问答的知识图谱网站演示,对农业信息化的发展具有重要意义。

关键词: 智能问答系统, 知识图谱, 双向长短期记忆模型(Bi-LSTM), 条件随机场(CRF)

Abstract:  Aiming at the difficulty of entity recognition in agricultural Q & A system, this paper proposed an improved Bi-LSTM-CRF based entity recognition method. Firstly, the word vector based on context information was generated through the pretreatment of BERT pretraining model, and then the trained word vector was input into Bi-LSTM-CRF for further training processing. Finally, the entity recognition, entity query and agricultural knowledge answering subsystems in the agricultural field were designed using Pythons Django framework. After experimental comparison, the improved Bi-LSTM-CRF proposed in this paper had better entity recognition ability in the field of agricultural information, and the precision rate, the recall rate, and F1 value on the agricultural information corpus were 93.23%, 91.08% and 92.16%, respectively. This paper realized the knowledge map website demonstration of entity recognition and agricultural information question answering in the agricultural field, which was of great significance to the development of agricultural informatization.

Key words: intelligent question answering system, knowledge graph, bidirectional long and shortterm memory model (Bi-LSTM), conditional random field (CRF)

中图分类号: