基于ChatGLM的水生动物疾病诊断智能对话系统的优化研究OA北大核心CSTPCD
Optimization of a ChatGLM-based intelligent dialogue system for aquatic animal disease diagnosis
水产品作为重要的食物来源之一,在养殖过程中出现的疾病问题严重影响着养殖业的可持续发展.针对水生动物疾病诊断智能对话系统存在复杂的专业性知识和准确性低的问题,提出一种基于ChatGLM模型的改进水生动物疾病诊断相关问题的优化方法.该方法通过在ChatGLM模型的中间层插入Adapter模块,针对相关的专业问题进行微调,提高了模型的专业性和准确性.同时采用P-tuning方法对输入部分进行高效的参数微调,使得对特定任务的调整更加精确.通过在水生动物疾病诊断对话数据集上的验证得出,该方法的双语评估替补(BLEU)指标从65.3%提升至75.1%,有效地解决了水生动物疾病诊断智能对话系统存在的准确性和专业性问题,为水生动物疾病诊断提供了有价值的辅助决策.
Aquatic products,as one of the important sources of food,are seriously affected by diseases that occur during the aquaculture process,which can seriously affect the sustainable development of the aquaculture industry.In allusion to the problems of complex professional knowledge and low accuracy in the intelligent dialogue system for diagnosing aquatic animal diseases,an optimization method based on the ChatGLM model is proposed to improve the diagnosis of aquatic animal diseases.In the method,the expertise and accuracy of the model are improved by inserting the Adapter module into the middle layer of the ChatGLM model to fine-tune the related specialized problems.The P-tuning method is used to efficiently fine tune the input parameters,making adjustments to specific tasks more precise.By the validation on the aquatic animal disease diagnosis dialog dataset,the BLEU(bilingual evaluation understudy)index of the method is improved from 65.3%to 75.1%,which can effectively improve the accuracy and professionalism of the aquatic animal disease diagnosis intelligent dialog system and provide valuable auxiliary decision-making for aquatic animal disease diagnosis.
尹娴;冯艳红;叶仕根
大连海洋大学 信息工程学院,辽宁 大连 116023大连海洋大学 水生动物医院,辽宁 大连 116023
电子信息工程
ChatGLM水生动物疾病诊断智能对话系统Adapter模块P-tuning方法BLEU
ChatGLMaquatic animal disease diagnosisintelligent dialogue systemAdapter moduleP-tuning methodBLEU
《现代电子技术》 2024 (014)
177-181 / 5
辽宁省教育厅基本科研项目(JL201917)
评论