Building Smart Libraries with Large Language Model:Frameworks and the Future

Abstract

Abstract:

This paper aims to explore the way of developing smart library services with the help ofrelated technologies in the era of large language model (LLM). It designs the system architecture of LLMthat works towards smart library services, analyzes the problems faced by libraries in the era of LLM,examines the application characteristics of LLM in diverse scenarios in the library, and demonstrates itsvalue and feasibility to libraries with the help of relevant cases. By putting forward suggestions on howto effectively transform libraries in the face of information changes brought by LLM, this paper aimsto promote the improvement of library service level, advocate the innovation and iteration of libraryknowledge services, and actively explore the application possibilities of LLM in the library.

Guo Limin(Shanghai Library) Fu Yaming (School of Cultural Heritage and Information Management, Shanghai University). Building Smart Libraries with Large Language Model:Frameworks and the Future[J]. Libraly Journal, 2023, 42(391): 22-30.

References

[ 1 ] 中_国通信院 . 人工智能生成内容（AIGC）白皮书（2022年）[EB/OL]. [2023-07-26]. http://www.caict.ac.cn/kxyj/qwfb/bps/202209/t20220902_408420.htm.

[ 2 ] 深圳市人工智能产业协会 . 一文读懂可信AI 大模型标准体系[EB/OL]. [2023-07-26]. https://szaicx.com/page142?article_id=13664.

[ 3 ] 张晓林，梁娜 . 知识的智慧化、智慧的场景化、智能的泛在化——探索智慧知识服务的逻辑框架[J]. 中_国图书馆学报，2023，49（3）：4–18.

[ 4 ] Panda S， Kaur N. Exploring the viability ofChatGPT as an alternative to traditional chatbotsystems in library and information centers[J].Library Hi Tech News，2023，40（3）：22–25.

[ 5 ] Chen X. ChatGPT and its possible impact onlibrary reference services[J]. Internet ReferenceServices Quarterly，2023，27（2）：121–129.

[ 6 ] Lund D B，Wang T. Chatting about ChatGPT： howmay AI and GPT impact academia and libraries?[J].Library Hi Tech News，2023，40（3）：26–29.

[ 7 ] 赵瑞雪，黄永文，马玮璐，等 . ChatGPT 对图书馆智能知识服务的启示与思考[J]. 农业图书情报学报，2023，35（1）：29–38.

[ 8 ] Murphy K P. Probabilistic machine learning： anintroduction[M]. MIT Press，2022.

[ 9 ] Gozalo-Brizuela R， Garrido-Merchan E C.ChatGPT is not all you need. A State of the ArtReview of large Generative AI models[J]. arXivpreprint arXiv：2301.04655，2023.

[10] 上海图书馆开放数据竞赛 . 梦中_奇缘[EB/OL].[2023-02-20]. http://wrd2016.library.sh.cn/channel/video/?ccid=607A3DE098DCE7329C33DC5901307461.

[11] MyHeritage. Deep Nostalgia[EB/OL]. [2023-02-22].https://www.myheritage.com/deep-nostalgia.

[12] 北京大学数字人文研究中_心 .《永乐大典》高清影像数据库系统[EB/OL]. [2023-07-27]. https://pkudh.org/project/yongledadian/.

[13] GitHub. AutoGPT[EB/OL]. [2023-07-20]. https://github.com/Significant-Gravitas/Auto-GPT.

[14] Wei J， Wang X， Schuurmans D， et al. Chainof-thought prompting elicits reasoning inlarge language models[J]. Advances in NeuralInformation Processing Systems，2022，35：24824–24837.

[15] 郭利敏_ . 基于卷积神经网络的文献自动分类研究[J].图书与情报，2017（6）：96–103.

[16] 郭利敏_，葛亮，刘悦如 . 卷积神经网络在古籍汉字识别中_的应用实践[J]. 图书馆论坛，2019，39（10）：142–148.

[17] 陈大庆 . 新时代图书馆数字化转型的思考[J]. 大学图书馆学报，2022，40（6）：14–16.

[18] Library of Congress. Newspaper Navigator[EB/OL].[2023-02-17]. https://labs.loc.gov/work/experiments/newspaper-navigator/.

[19] Zhu Y， Wan J， Zhou Z，et al. Triple-to-text： convertingRDF triples into high-quality natural languagesvia optimizing an inverse KL divergence[C]//Proceedings of the 42nd International ACM SIGIRConference on Research and Development inInformation Retrieval. 2019：455–464.

[20] Distiawan B， Qi J， Zhang_ R， et al. GTR-LSTM：A triple encoder for sentence g_eneration from RDFdata[C]//Proceeding_s of the 56th Annual Meeting_of the Association for Computational Ling_uistics（Volume 1： Long_ Papers） . 2018：1627–1637.

[21] Bommasani R， Hudson D A， Adeli E， et al. Onthe opportunities and risks of foundation models[J].arXiv preprint arXiv：2108. 07258，2021.