分类： 计算机科学 >> 计算机应用技术 分类： 图书馆学、情报学 >> 情报学 提交时间： 2023-05-12
摘要： 目的/意义：随着以ChatGPT为代表大语言模型技术的不断发展与变革，使得许多领域的经典场景都重新焕发出新的机会。同时，越来越多的学者开始关注如何将大语言模型的智能化能力与技术应用到现有的场景，并分析这些技术带来的挑战和机遇。 方法/过程：本文以ChatGPT为建模对象，首次将大语言模型技术引入用户图书评分偏好预测这一图情领域的典型应用场景，并落地实践。通过构建基于ChatGPT的用户图书评分预测模型（CUBR, ChatGPT-based model for User Book Rating Prediction），来探索大语言模型技术在图书推荐领域实践和落地的可行性。同时，本文基于图书评分任务的不同评估方案与现有经典推荐模型进行对比，探讨并给出了CUBR在用户图书评分预测场景的优势与劣势，并分析了后续大语言模型在图书推荐其他场景可能的研究机会点。 结果/结论：本文实验研究表明，（1）CUBR模型在现有用户图书评分偏好预测任务上能够取得不错的推荐效果，特别是单样本(One-shot)这类待推荐目标信息较少的情况下，其表现接近或超过当前经典推荐算法，且泛化能力较强，较适用于冷启动推荐场景。（2）随着单个用户提示样本内容的增加（如从One-shot到Ten-shot），CUBR的预估效果会有显著的提升，说明CUBR具备不错的实时上下文学习能力。 局限：本文研究场景仅限于用户图书评分偏好理解与推荐，未来将尝试在更多的图情场景应用和改造现有大语言模型技术，并获得更好的实践效果。
分类： 图书馆学、情报学 >> 图书馆学 分类： 计算机科学 >> 计算机应用技术 提交时间： 2016-04-12
摘要： Open Access academic paper has become important measures of the world's leading countries which promote knowledge sharing, collaborative open innovation, economic growth and inclusive development. Moreover, Open Sharing of academic papers founded by public funding projects has become the consensus of the world’s leading countries, and an important requirement is to deposit those papers in the open access IRs which are attribute to funder and authors’ institute. But the situation of institutional repository in China is serious, such as poorer deposit awareness, incomplete or incorrect submitted data and so on. It affects the development and perfection of China Open Access and open sharing mechanism. Though RJ-Broker can assist and promote to solve problems above, it is mainly related to European PMC data and located abroad. In order to better solve those problems, National Science Library of CAS, as a leader institute of this study in China, with hundreds of Research institutes of Chinese Academy of Sciences as demonstration, and reference to the OA-RJ model, constructed a pushing and routing services of Chinese academic papers, namely iSwitch, implementing automatic deposit. It can help institutes and funders to construct their IRs effectively and promote academic papers utilization by others openly. After iSwitch service is released publicly, it has routed more than 360,000 paper metadata pushed by Web of Science and some experimental Open Access paper data from other publishers to CAS IRs. Now it is a stable service to exchange WOS update data and other publisher data. Besides, with the help of iSwitch, Web of Science has linked full-text link of CAS IR papers since 27, July, 2015.
点击量
待评议
点击量
下载量
评论
