融合情感信息的开放域对话生成算法在学前教育聊天机器人中的应用研究  

Research on the Application of an Open Domain Conversation Generation Algorithm Integrating Emotional Information in a Chat Robot for Preschool Education

作　　者：黄玉芳[1] HUANG Yufang(Xi’an Fanyi University,Xi’an 710100,China)

机构地区：[1]西安翻译学院,西安710100

出　　处：《自动化与仪器仪表》2024年第7期266-269,共4页Automation & Instrumentation

基　　金：陕西省教育科学规划2023年度课题《满意度的视域下应用型民办本科院校学前教育专业教材质量评价标准研究》;西安翻译学院2023年度校级科研基金《<评估指南>背景下幼儿教师情感支持质量研究——基于CLASS课堂互动评估系统的观察分析》(23B06)。

摘　　要：为进一步提升学习教育机器人的教育质量,提出一种基于改进情感监督的对话生成模型。其中,以双向长短时记忆网络Bi-LSTM作为基础的对话生成方法,将其与情感监督的方法相结合,以进一步提升生成对话回复的质量。实验结果表明,与传统的端到端模型以及未改进的情感监督模型相比,改进的情感监督模型所生成的对话回复的质量明显更高,流畅度更高,更加自然并贴合人们实际的对话,在bigram多样性评价指标上具有明显的优势。综上,设计的基于改进情感监督的对话生成模型性能良好,能够生成质量较高的对话回复,将其应用于学前教育的场景中时,能够帮助儿童进行更高质量的对话学习。To further improve the educational quality of learning and educational robots,a dialogue generation model based on improved emotional supervision is proposed.Among them,the dialogue generation method based on the bidirectional long-term and short-term memory network Bi-LSTM is combined with the method of emotional supervision to further improve the quality of dialogue responses produced.The experimental results show that compared with traditional end-to-end models and unimproved emotional supervision models,the improved emotional supervision model generates significantly higher quality,smoother,more natural,and tailored to people's actual conversations.It has significant advantages in evaluating the diversity of bigram indicators.In summary,the designed dialogue generation model based on improved emotional supervision has good performance and can generate high-quality dialogue responses.When applied in preschool education scenarios,it can help children engage in higher quality dialogue learning.

关 键 词：学前教育 对话生成 情感分析 Bi-LSTM 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]