Samuel Hamman W, Zaïane Osmar R
Department of Computing Science, University of Alberta, Edmonton, Alberta, Canada.
Stud Health Technol Inform. 2017;245:207-211.
We present a recommender system, PubMedReco, for real-time suggestions of medical articles from PubMed, a database of over 23 million medical citations. PubMedReco can recommend medical article citations while users are conversing in a synchronous communication environment such as a chat room. Normally, users would have to leave their chat interface to open a new web browser window, and formulate an appropriate search query to retrieve relevant results. PubMedReco automatically generates the search query and shows relevant citations within the same integrated user interface. PubMedReco analyzes relevant keywords associated with the conversation and uses them to search for relevant citations using the PubMed E-utilities programming interface. Our contributions include improvements to the user experience for searching PubMed from within health forums and chat rooms, and a machine learning model for identifying relevant keywords. We demonstrate the feasibility of PubMedReco using BMJ's Doc2Doc forum discussions.
我们展示了一种推荐系统——PubMedReco，用于从拥有超过2300万条医学引用文献的数据库PubMed中实时推荐医学文章。PubMedReco能够在用户于诸如聊天室这样的同步通信环境中交谈时推荐医学文章引用文献。通常情况下，用户不得不离开聊天界面去打开一个新的网页浏览器窗口，并制定一个合适的搜索查询来检索相关结果。PubMedReco会自动生成搜索查询，并在同一个集成用户界面中显示相关引用文献。PubMedReco分析与对话相关的关键词，并使用它们通过PubMed电子工具编程接口搜索相关引用文献。我们的贡献包括改善在健康论坛和聊天室中搜索PubMed的用户体验，以及一个用于识别相关关键词的机器学习模型。我们使用英国医学杂志的Doc2Doc论坛讨论来证明PubMedReco的可行性。
