摘要
[目的/意义]智慧服务已成为高校图书馆转型发展的重要方向。自然语言处理技术赋能了高校图书馆智慧化服务,带来了服务模式与流程的重构,有助于提升高校图书馆的整体服务水平。命名实体识别是自然语言处理中的一个重要任务,对图书馆智慧化服务产生重要影响和价值,可有效识别信息中的人名、地名、组织机构、资源利用、服务特色和文化推广等实体,为知识组织、信息检索等提供支持。[方法/过程]本文利用命名实体识别技术分析高校图书馆智慧化服务系统的应用前景,通过构建高质量的高校图情语料库,提供高质量训练数据,以满足领域内特定的实体识别需求的准确性和适应性,为优化图书馆智慧服务系统提供基础。采用基于深度学习的ALBERT-BILSTM-CRF模型,用以验证命名实体识别任务的效果,将该模型应用于高校图书馆服务推荐和知识图谱的案例分析,并与现有的国内外主流大语言模型进行了效果比较。[结果/结论]结果表明,本文提出的方法有效地提高了高校图情领域命名实体识别的性能,有助于实现图书馆智慧化服务的推广与应用,同时也减少了资源浪费和训练成本。此外,本文还探讨了服务于图书馆领域大语言模型LibraryGPT的可能性,以便对未来高校图书馆智慧服务的推广和发展提供参考和借鉴。
[Purpose/Significance]Intelligent services have become an important direction for the transformation and development of university libraries.Natural language processing technologies empower the intellectualization of university library services,reconstructing service models and processes,and helping improve the overall service levels of university libraries.Named entity recognition is an important task in natural language processing,its impacts and significance on the intellectualization of library services can effectively identify entities such as people,locations,organizations,resource utilization,service features,and cultural promotion in library information,providing support for knowledge organization,information retrieval,etc.[Method/Process]This study analyzed the application prospects of named entity recognition technology in the intelligent service systems of university libraries.By constructing a high-quality library and information science corpus,it provided high-quality training data to meet the specific entity recognition needs within the field,enhancing accuracy and adaptability,and laying the foundation for optimizing library intelligent service systems.An ALBERT-BiLSTM-CRF model based on deep learning was adopted to verify the effects of named entity recognition tasks.Case studies on service recommendations and knowledge graphs in university libraries were performed and compared with existing mainstream large language models at home and abroad.[Result/Conclusion]The results show that the proposed method effectively improves the performance of named entity recognition in the university library information field,which is conducive to the promotion and application of intellectualized library services,while also reducing resource waste and training costs.In addition,this paper explores the possibility of LibraryGPT,a large language model serving the library field,in order to provide references and inspiration for the promotion and development of smart services in university libraries in the future.
作者
刘思得
李东升
Liu Side;Li Dongsheng(Library,Fujian Normal University,Fuzhou 350007,China;Library,Fujian University of Technology,Fuzhou 350118,China)
出处
《现代情报》
CSSCI
北大核心
2024年第12期102-121,共20页
Journal of Modern Information
基金
教育部产学合作协同育人项目“高校图书馆嵌入式信息素养教育实践研究”(项目编号:202102654006)
教育部人文社会科学规划基金项目“海峡两岸图书馆标准规范体系语义互联及知识库构建研究”(项目编号:20YJA870001)
福建省自然科学基金项目“基于预训练语言模型的公共事件UGC资源知识组织研究”(项目编号:2020J01889)
教育部产学合作协同育人项目“新时期文献信息资源保障与服务利用”(项目编号:231005469193731)。
关键词
高校图书馆
智慧服务
命名实体识别
大语言模型
university library
intelligence service
name entity recognition
large language model
