Several databases can now be searched using natural language
The ability to search using natural language and semantic search instead of having to use Boolean operators (AND, OR), phrase search, and truncation has been an argument used in recent years to sell AI search tools such as Elicit, SciSpace, and—for that matter—ChatGPT. Now, even established databases are beginning to offer the possibility to search with natural language.
The University Library has so far been reluctant to pay extra for these services, as we have assessed that the quality is not yet high enough, but some databases now include certain natural language searches in the packages we already pay for. However, AI-generated summaries or interpretations of articles are not included.
In Ebsco's databases (e.g., Cinahl, Education Research Complete, Business Source Premier, MLA, GreenFILE) and Web of Science, it is now possible to use Natural Language Search (Ebsco) or SmartSearch (WoS). Ebsco still has the classic search as standard—you must activate Natural Language Search yourself. In Web of Science, SmartSearch is the default, which means that you must actively opt out if you want to use traditional search.
Since semantic search means that the system tries to interpret the meaning behind the words, not just match exact terms, it can be useful for quickly getting an overview of relevant articles, especially in areas where you are unsure of the terms used. However, since you have no control over which search terms are used, natural language searches are not suitable for searches where you want comprehensive and transparent results. It can also produce more irrelevant hits than a search you build yourself.
We recommend that you try both types and evaluate the results you get, both in terms of the number of hits and how the hits are ranked, and of course in terms of how relevant the hits are. Also note that in the Ebsco databases, you can see which search string the database has used, and in Web of Science, you can choose whether the results should come from a Boolean or semantic search, or both.
Are you a research group that would like a more detailed explanation of this? Contact forskarservice@ub.gu.se
Read more about natural language searches in librarian Aaron Tay's Musings about Librarianship:
About natural language search in Ebsco databases:
A Deep Dive into EBSCOhost's Natural Language Search and Web of Science Smart Search - Two bundled "Ai-powered"search (I)
About SmartSearch in Web of Science:
A Deep Dive into EBSCOhost's Natural Language Search and Web of Science Smart Search - Two bundled "Ai-powered"search (II)