Show me your search strings!
What can search statistics tell us about user behaviour? Librarians at the University of Tromsø, Norway analysed over 20,000 searches to find answers.
We are all expert searchers!
Powerful search engines make sure we get the answers we need. Or think we need. Or at least something close to what we think we need. The fact that we are able to get a long result list on whatever topic we search for does something to us and our expectations of what technology can help us with. We feel confident because we managed to find an answer. What happens when this confidence is transferred to a library database?
In an attempt to find out more about user search behaviour we looked at more than 21,000 searches done in 2012 in our previous federated search engine. Our findings were eye-opening. Some searches were quite good. Almost one in five searches yielded fewer than 100 results. We concentrated on the remaining 80%. The most disturbing finding was that more than 60% of all the searches ended up with no hits at all. We suspect some of these were timed-out searches, i.e. the search took too long to do, a familiar problem with our federated search option. But even if we took away these, there was still a vast amount of searches that resulted in no answer at all for the user
Spelling is not always easy, especially when it is in a language not your own. Search engines, and good discovery systems, typically try to fix your spelling error by asking "Did you mean …?" A traditional library database often doesn’t provide this service. We found several examples of this: canser, occupational, sergery, psycological are just a few illustrations. Another very common mistake is searching for very broad topics like history, therapy or fish, thus getting thousands of results.
With a myriad of databases to choose from, it’s no wonder library users sometimes get it wrong. We found searches for 'Dorian Gray' in Early English Books Online and we found a search for 'Wayne Rooney' in Psychbooks. A wrong use of search fields was also frequent among those who had ventured into advanced search. Especially the subject field seems to confuse users. We found searches like 'speed of processing' and 'age-related' as well as whole article titles entered into the subject field. The most surprising mistake was however the wrong use of Boolean operators. What is pure logic to any librarian is apparently not that clear to everyone. We found searches like 'nursing OR hypochondriasis' and 'practice knowledge OR health'. Similarly we found the same word in two different languages combined with and, like 'role-play AND rollelek', rollelek being the Norwegian word for role-play.