From search words to search questions?

Are you good at asking questions but not a fan of ANDs and ORs and brackets? Then you might like what is happening with the search engines.

Information retrieval instructions traditionally start by finding out the keywords of your topic and combining them into a search query. Then, synonyms and other related terms are considered and added to the search query using Boolean operators (AND, OR). This is still the most effective method, for example, when conducting information retrieval through SAMK-Finna. Finna doesn’t yet try to think on behalf of the searcher. Not yet.

Google, on the other hand, has been trying to think on behalf of the searcher for a long time – often successfully. The same goes for its scholarly version, Google Scholar. Up until now, I have believed that it’s worth starting searches there with simple keyword combinations and then progressing to more advanced search queries. However, now it might be time to approach information retrieval also a bit differently.

Let’s do a little test with the topic of “treating knee arthritis without surgery”. It’s a good topic to test a search engine’s natural language understanding because it includes the concept of “without surgery”. If the search engine understands natural language, it should be able to retrieve studies on treatment methods other than surgery. If it doesn’t understand natural language, it may miss the meaning of the word “without” and search only for the term “surgery”, providing especially surgical treatment-related studies.

If you search directly with the phrase treating knee arthritis without surgery, you’ll find results related to treatment without surgery, but also results related to surgery. Pretty good, but it could be better.

Let’s change the search query to the form of a question: How can knee arthritis be treated without surgery? The question doesn’t use the most scientifically precise vocabulary, but still, all the articles at the top of the result list deal with treatment without surgery. They use terms like “non-surgical”, “nonoperatively”, “non-arthroplasty” etc. instead of “without surgery.” Even a simple question posed in natural language helped Google Scholar figure the type of information we were looking for. The search results can be refined by refining the questions: What are the best evidence-based non-surgical treatment methods for knee osteoarthritis? This is absolutely worth experimenting more.

Google Scholar is, in that sense, a traditional search engine in that it provides a list of sources as a result (at least for now). In addition, there are now search engines available that also respond to questions based on research. In our example case, they would tell you about treatment options and what studies say about them. For instance, Consensus is one free option (requires logging in). As paid options, you can get better ones like ChatGPT Plus extensions.

In the future, will it be more important to know how to ask the right questions from a search engine than to rely on keywords and their combinations? The question doesn’t have to be spot on immediately — getting started with something sufficiently good is enough. Perhaps in the future, we will begin our searches with well-formed questions and then continue with traditional search techniques if needed. However, the fundamental principle remains that you must be able to express your information needs clearly and concisely — one way or another.

P.S. If you haven’t linked Google Scholar results to SAMK Library yet, please do so. You’ll get access to a lot more full texts from your search results. This video explains how to do it.

Text: Teppo Hjelt, Information Specialist at SAMK Library, who works on guidance related to information retrieval
Image: Teppo Hjelt, created with DALL-E 3 AI