摘要
中华古典服饰作为中国传统文化的重要组成部分,不仅是历史文化的载体和传承者,还深刻反映了社会等级与身份地位,有力地促进了文化认同与传统价值的传播与延续。然而,由于历史变迁、文化演进以及现代生活方式的转变,许多人对中华古典服饰的了解较为有限。为弥补这一知识空缺,开展了中华古典服饰知识图谱的研究与构建,并在此基础上进一步创立了智能问答系统。针对用户输入的问题,系统采用了一种基于BiLSTM模型的改进模型RoBERTa_BiLSTM_SDPA_CRF,对问句进行命名实体识别,再利用RoBERTa对问句进行分类,并引入新的召回排序模块计算相似度,最终输出对应的答案。在中华古典服饰问句数据集的测试中,改进模型在精确率、召回率和F1值等关键指标上均优于其他模型。改进的整体问答结构与以前的问答结构相比,在两个数据集的F1值分别提升了约2.73%和4.20%。结果表明,构建的问答系统在中华古典服饰知识的检索和利用上取得了良好的应用效果。
Classical Chinese costume play an essential role in traditional Chinese culture,they are not only the carrier and inheritance of history and culture,but also reflect social hierarchy and status,and promote the dissemination and continuation of cultural identity and traditional values.However,due to changes in history,culture and modern lifestyles,many people have a limited understanding of classical Chinese costumes.To bridge this gap,a comprehensive knowledge graphs of traditional Chinese costumes was conducted and an intelligent question answering system based on knowledge graphs was developed.For the question queried by user,an improved model RoBERTa_BiLSTM_SDPA_CRF based on BiLSTM model is used for named entity recognition of the interrogative sentence,and then ReBERTa is used to classify the interrogative sentence,and a new recall sort module is added to calculate the similarity,and finally the corresponding answer is output.In the dataset of Chinese classical Costume questionnaire,the improved model RoBERTa_BiLSTM_SDPA_CRF model is improved in terms of precision,recall and F1 compared with other models.In pairs of the improved structure of Q&A compared with the previous structure of Q&A,F1 was improved by about 2.73% and 4.20%,respectively.The constructed Q&A system has achieved good application results in the retrieval and utilisation of classical Chinese Costume knowledge.
作者
李瑜涵
张俊杰
袁桦
LI Yuhan;ZHANG Junjie;YUAN Hua(Computer and Artificial Intelligence College,Wuhan Textile University;Hubei Provincial Engineering Research Center for Intelligent Textile and Fashion;School of Fashion,Wuhan Textile University;Hubei Research Center of Fashion Art and Culture,Wuhan 430073,China)
出处
《软件导刊》
2025年第8期145-151,共7页
Software Guide
基金
湖北省教育厅哲学社科基金一般项目(23Y151)。
关键词
知识图谱
问答系统
命名实体识别
中华古典服饰
knowledge graph
question and answer system
named entity recognition
traditional Chinese classical costume
作者简介
李瑜涵(2001-),女,武汉纺织大学计算机与人工智能学院硕士研究生,研究方向为问答系统、知识图谱;张俊杰(1980-),男,博士,纺织服装智能化湖北省工程研究中心副教授,研究方向为知识图谱、推荐系统;通讯作者:袁桦(1982-),女,博士,武汉纺织大学服装学院讲师,研究方向为时尚设计与纺织服装行业可持续发展。
