Wednesday August 30, 2006
Search results: paging, relevance and useability
After reading a few blog entries some time back, I wrote down a couple of ideas for improving the useability of search results.
The blog entries that kicked it off are below:
A comment on the SvN blog by someone using the handle Mahjong Montréal has hit upon something here: we are busy discussing the best way to resolve the symptoms of unuseable search results rather than tackling the root problem.
The problem at hand is not the pagination of search results, but rather: How can we present search results in useable (and reuseable) fashion. The the constraints on any solutions are defined by a human’s ability to transport or store said information, view it comfortably or process it in short-term memory.
It is reasonable to expect that the solution will differ for different types of information, and we see this already: chapters and page numbers for books, site maps for websites, etc.
Pages are great when:
- the information being broken up is more-or-less ‘linear’ and follows a coherent path
- the transition between pages does not unduely interrupt the processing of the information (compare the time taken to turning the page in a book to loading a new page of search results at a search engine)
- non-semantic symbols (e.g. page numbers) can aide in finding ones place to contine the ‘linear’ communication
Search results are typically a collection of options rather than a coherent ‘linear’ communication. As one commenter pointed out, what is needed is the ability to increase the relevance and accuracy of search results, and thus reduce the result set to precisely and only those options that are relevant. This would eliminate the need for paging because it then becomes desirable to have this carefully selected group of options presented together.
Such a goal is beyond the scope of a user interface. Real advances in seach relevance will come from better algorithms that understand intent and scope rather than just words. What is required in the meantime is a way of allowing the searcher to reduce the search results to a list of options that are relevant to him.
iht.com allows visitors to click an icon beside the headline of any story to add it to a personalised list of ‘Clippings’. This clippings list acts as a store of what is relevant to that visitor, who can then visit each ‘clipped’ story at leisure.
This would be ideal for search engine results: let the searcher pick which of the proffered results he is interested in further exploring. By ticking these off, create a list of options that is relevant and no one longer has to worry about page numbers, endless scrolling and relative positions to find any particular search result again.
A9 goes part way by saving searches and which results I clicked through on; but what is needed is to save the search along with the options that I would mark as be worth pursuing and then perhaps track if those options have been visited or not.
How can this be done? Do we pick relevant links to form our list or do we knock irrelevant items out of the initial search results?
What do you think?
As an aside to this, it occured to me that once we have only search results that are relevant to our purpose, they can easily be tagged and saved using the search query to provide the initial tags.
Filed by Jachin Sheehy under