中文题名: | 面向中学语文课文的知识图谱构建与应用研究 |
姓名: | |
保密级别: | 公开 |
论文语种: | 中文 |
学科代码: | 081203 |
学科专业: | |
学生类型: | 硕士 |
学位: | 工学硕士 |
学位类型: | |
学位年度: | 2021 |
校区: | |
学院: | |
研究方向: | 自然语言处理 |
第一导师姓名: | |
第一导师单位: | |
提交日期: | 2021-06-19 |
答辩日期: | 2021-06-19 |
外文题名: | RESEARCH ON THE CONSTRUCTION AND APPLICATION OF KNOWLEDGE GRAGH FOR MIDDLE SCHOOL CHINESE TEXT |
中文关键词: | |
外文关键词: | Knowledge Graph ; Middle School Chinese learning ; Knowledge extraction ; Representation learning ; Three-way decision ; Knowledge Graph Visualization |
中文摘要: |
随着语文教学改革在中学的全面推行,语文学科的阅读内容和阅读难度均大幅增加,全面提升中学生的阅读能力迫在眉睫。一方面,中学是提高学生语文知识积累和提升学生学习与阅读能力的关键阶段。另一方面,随着人工智能领域的发展,如何通过科技手段进行有效学习越来越受到人们的关注。知识图谱作为一种人工智能知识引擎,以其高效存储与数据处理的能力突破了传统知识库的局限性,为系统梳理语文知识和实现阅读内容推荐提供了可能。 本文从语文教材课文出发,通过广泛收集相关资源,针对中学语文课文构建了一个中等规模的知识图谱。以此为基础,采用翻译模型对知识图谱进行知识表示,并提出了基于三支决策的阅读文本推荐模型。进而,将研究结果应用于“素养植根语文融媒体教材”APP中,为语文教学提供辅助服务。具体来说,本文工作及创新主要包括: (1) 中学语文课文知识图谱的构建与可视化。首先,构建了中学语文知识图谱本体库,其中包括15个实体类、44个属性、22种关系。进而,使用网络爬虫和自然语言处理相关技术获取教材课文及赏析资源,并设计了半自动化的信息抽取方法实现对不同类型数据源的知识抽取,最终得到16442条知识三元组。最后,采用Neo4j图数据库实现中学语文知识图谱的可视化,完成中学语文知识图谱的构建工作。 (2) 基于知识表示学习的阅读文本推荐。首先,通过对比Trans系列模型在本文知识图谱上的链接预测实验效果,得到最佳模型及参数设置。进一步的,为将模型更好地应用于中学语文知识图谱的语义向量表示,为知识点设定了重要度规范,重组后的数据链接预测结果表明,改进后的方法对本文构建的知识图谱具有更好的表示效果。进而,基于优化后的知识图谱向量表示,提出了基于三支决策的阅读文本推荐模型,结合相似度计算结果实现了阅读文本推荐功能。经专家验证,推荐结果相对合理,具有较强的教学参考性。 中学语文知识图式学习系统设计。将中学语文课文知识图谱及阅读文本推荐方法应用到语文教学APP,设计面向语文教学APP前端的知识图谱可视化系统,作为教师开展课堂教学和学生自主学习的辅助手段,促进用户对语文学科的知识认知,提供语文教学的知识库与搜索引擎服务。 |
外文摘要: |
With the comprehensive implementation of the Chinese teaching reform in middle schools, the reading content and reading difficulty of Chinese subjects have increased significantly, and it is urgent to comprehensively improve the reading ability of middle school students. On the one hand, middle school is a key stage to improve the accumulation of students' Chinese knowledge and improve their learning and reading ability. On the other hand, how to conduct effective learning through scientific and technological means has attracted more and more attention. As an artificial intelligence knowledge engine, the knowledge graph breaks through the limitations of traditional knowledge bases with its ability to store and process data efficiently, and provides the possibility for systematically combing Chinese knowledge and implementing reading content recommendation. Starting from the texts of Chinese textbooks, this dissertation constructs a middle-scale knowledge map for middle school Chinese texts through extensive collection of related resources. Based on this, the translation model is used to represent the knowledge graph, and a reading text recommendation model based on three decisions is proposed. Furthermore, the research results are applied to the APP of "Quality-rooted Chinese Language and Media Teaching Material" to provide auxiliary services for Chinese teaching. Specifically, the work and innovations of this dissertation mainly include: (1) The construction and visualization of the knowledge map of middle school Chinese texts. Firstly, the ontology database of middle school Chinese knowledge graph is constructed, which includes 15 entity classes, 44 attributes, and 22 relationships. Furthermore, it uses web crawlers and natural language processing related technologies to obtain textbooks, texts and appreciation resources, and designs a semi-automatic information extraction method to extract knowledge from different types of data sources, and finally 16442 knowledge triples are obtained. Finally, the Neo4j graph database is used to realize the visualization of the middle school Chinese knowledge graph and complete the construction of the middle school Chinese knowledge graph. (2) Reading text recommendation based on knowledge representation learning. First, by comparing the link prediction effect of the Trans series model on the knowledge graph of this article, the best model and parameter settings are obtained. Furthermore, in order to better apply the model to the semantic vector representation of the middle school Chinese knowledge graph, the importance of the knowledge points is set, and the data link prediction results after reorganization show that the improved method has a better presentation effect on the knowledge graph constructed in this dissertation. Furthermore, based on the optimized knowledge graph vector representation, a reading text recommendation model based on three decisions is proposed, and the reading text recommendation function is realized by combining the similarity calculation results. Verified by experts, the recommended results are relatively reasonable and have strong teaching reference. (3) The design of the schema learning system of middle school Chinese knowledge. Apply the knowledge map and reading text recommendation method of middle school Chinese texts to the Chinese teaching APP, design a knowledge map visualization system for the front end of the Chinese teaching APP, as an auxiliary means for teachers to carry out classroom teaching and students' independent learning, promote users' cognition of Chinese subjects, and provide knowledge base and search for Chinese teaching Engine service. |
参考文献总数: | 46 |
作者简介: | 人工智能学院2018级硕士研究生 |
馆藏号: | 硕081203/21003 |
开放日期: | 2022-06-19 |