Focused Clinical Search through Query Intent Interpretation and a Healthcare Knowledge Graph
Posted: 29 Jan 2021
Date Written: November 5, 2020
Clinicians and medical professionals need accurate, succinct, updated, and trustworthy snippets from medical literature as responses to search queries made during the diagnosis, prognosis, or treatment of any patient. Barriers to focused clinical search often include the insufficient search time, need to search for patient comorbidities and contexts, the ever-growing volume of medical evidence, lack of awareness of which resource to search for specialty questions, and skepticism and lack of trust regarding the quality of search results. In this talk, we will be presenting research and development behind a production-ready Focused Clinical Search Service (HGFCSS), which is powered by Elsevier’s Healthcare Knowledge Graph (HG), to facilitate the retrieval of relevant medical content for clinical search queries made at point of care. HG contains medical knowledge (concepts, relations, cohorts, etc.), regularly curated by subject matter experts, and novel clinical relations, extracted by natural language processing (NLP) models from unstructured medical content through automated pipelines.
Given a clinical search query, the HGFCSS identifies the core HG medical concepts as well as additional refinements proposed by the user (e.g., treatment, diagnosis) in that query. The service intelligently infers the clinical query intent as well as performs corrections on misspelled words. Query refinements are further interpreted with respect to structural elements in literature sources and HG relation types. Using a federated querying infrastructure over multiple search indexes, we can retrieve the right section in the right chapter from a diverse set of literatures sources, such as reference medical textbooks, synoptic curated clinical content, and drug monographs, for focused clinical search queries with a desirable query response time. Using automation methods and regular update of concepts, labels, and relations from several sources, we can ensure that clinicians are always equipped to find recent, succinct, and trustworthy medical content for their queries (e.g., covid diagnosis, temozolomide indications).
The novel HGFCSS is effective compared to Elsevier’s popular Clinical Key Search Engine to retrieve relevant content excerpts for focused clinical search queries, as measured by DCG and other metrics. We are planning to extend the HGFCSS’s query parsing and information retrieval methods to use word embeddings and learning to rank models to improve the query intent interpretation, to retrieve more related content, and to improve ranking of the search results.
Keywords: Knowledge Graph, Clinical Search, Information Retrieval, Natural Language Processing, Automation
Suggested Citation: Suggested Citation