Personalized Search

In reading The Search one of the topics that comes up is the goal of the perfect search. In order to achieve the perfect search it’s believed you need to have personalized search so that searches can be put into context and the likely search “intent” of the searcher. A person that searches for coffee a lot and a person who searches are mostly computer related may have a different intent of “java”. Right now the searcher is required to add more context to the item if they want good results but the goal is too move most of this to the backend.

Since I use Google a lot I went ahead and turned on their personlized search. While this is just the beginning, and doesn’t do the above, the interface is quite interesting and may give some ideas of what could be done in the OPAC. I’m not a huge fan of the whole saved records set-up, email, etc that some OPACs and online databases have but at least it’s something. I don’t remember when I turned on the search in google but it appears to have been sometime in December.

Personalized Search: Trends

Google Personal Search Trends

As you can see above, Google breaks down my searches by date, time and frequency. You can click on the dates as well and see the top searches for that month, day or hour. While this looks like it would be more cool factor than useful I’ve already found it worthwhile. I had a search the other day where I went through many pages of results as I just couldn’t find the right keyword. I eventually found a page that was worthwhile though I forgot to bookmark it. No problem as Google now tells me which results I clicked on.

Personal Search: Bookmarks

Google Personal Search Bookmarks

If the trends didn’t catch your fancy, you may be interested in the bookmark feature. When you click on searches you can see your history and then “star” things you want to save. You can then also apply labels to it. For those with Gmail your already familiar with this interface. I can only presume that bookmarks will be integrated with gmail sometime in the future.

An interesting note: It tells me what my failed searches were. An interesting tidbit of data and something I may use in the future to help me refine searches.

Uses of the Data for OPACs

Setting aside the many privacy issues, what can this data be useful for if similar things were done in an OPAC environment. The first thing I can think of is suggestions. When you know what searches work and which ones fail then building smart suggestions for users becomes much more useful. Right now I’ve seen various implementations of suggested searches which are based on such things as whether there are results, etc. This is good in it’s own right as it prevents suggestions that don’t return anything, but it would be even better when you can bump up suggestions that have a high success ratio. This would be easier to create, presuming your library has holds, etc. If it doesn’t then it may be difficult to track “success” other than a specific item was viewed.

Another possibility is a smart citation list. The ability to keep and tag citations as well as the searches that got them adds another powerful layer to research. Some of this is already possible using external services but as most OPACs have some sort of “mark record” functionality it would be nice to have a more versatile system. Again, a nice API would be helpful here so that this information can be moved in and out of the system. The ability to organize research by project, go back to “good” searches if not enough information ended up being available and seeing what searches came up empty would all be helpful in this process.

There are probably a hundred other things that could be done on the server side that the patron never sees but these are just a few that can help on the public face of things. If you don’t mind your searches being logged to your account then I recommend giving it a whirl.

One Response to “Personalized Search”

  • 1
    calebtr Says:

    re: opac applications, and not completely on-topic, i think it’s long overdue that we (try to) use search data to help out our relevancy rankings in keyword searches, except instead of tracking each individual’s personalized searches, track the community of people who search.

    privacy-wise, an anonymous session cookie could record the search and rate success based on what records were viewed (some points), how many records were viewed for that session-search (a search is less helpful if you have to look at lots of records), and whether or not the link to place a hold was followed from any record (lots of points).

    later on, the next searcher benefits.

    librarians can even seed rankings (call it cataloging).