There is an interesting note in Jimmy Wales article:
“Our Wikianswers Q&A site in particular has seen tremendous growth since it was re-launched in February, and we’ll be investing more of our time in it.”
Compared to the failure of Wikia search to take off, it seems that people actually enjoy answering other people’s question, but not so much reordering their search results. Maybe it is because the task of result re-ordering is too abstract and artificial, while question answering is something we do naturally.
To take things one step further: what is needed then to make user feedback (explicit, like results reordering, not implicit like clicks) more enjoyable for users, and as a result more useful for search?
Maybe that’s what the Hunch guys and Luis Von Ahn are after–turn human computation into a game.
But I think Wikia Search and Google’s Search Wiki are making the same mistake: Wikipedia works because people are editing pages. That’s a lot more meaningful –and yes, more concrete–than editing a search result list.
I think relevance feedback makes sense when it fits naturally into the information seeking process. People do use query refinement interfaces, like faceted search. Also, if you ignore the problem of gaming, then clickthrough data provides implicit relevance feedback–and you could randomize result ordering in order to get a larger variety of such feedback (though you still have to calibrate to the base click-though rates associated with results at each rank). But I know there are lots of challenges using such data, so perhaps there is a need for explicit feedback mechanisms,
The best way to get feedback is to offer something of value to the user: help in fulfilling their immediate information-seeking task. If a refinement option, “more like this” button or any other mechanism helps users do so, they will pick up on it. Lazy, opportunistic users are our friends!
The Noisy Channel 
2 replies on “Wikia Search, R. I. P.”
There is an interesting note in Jimmy Wales article:
“Our Wikianswers Q&A site in particular has seen tremendous growth since it was re-launched in February, and we’ll be investing more of our time in it.”
Compared to the failure of Wikia search to take off, it seems that people actually enjoy answering other people’s question, but not so much reordering their search results. Maybe it is because the task of result re-ordering is too abstract and artificial, while question answering is something we do naturally.
To take things one step further: what is needed then to make user feedback (explicit, like results reordering, not implicit like clicks) more enjoyable for users, and as a result more useful for search?
LikeLike
Maybe that’s what the Hunch guys and Luis Von Ahn are after–turn human computation into a game.
But I think Wikia Search and Google’s Search Wiki are making the same mistake: Wikipedia works because people are editing pages. That’s a lot more meaningful –and yes, more concrete–than editing a search result list.
I think relevance feedback makes sense when it fits naturally into the information seeking process. People do use query refinement interfaces, like faceted search. Also, if you ignore the problem of gaming, then clickthrough data provides implicit relevance feedback–and you could randomize result ordering in order to get a larger variety of such feedback (though you still have to calibrate to the base click-though rates associated with results at each rank). But I know there are lots of challenges using such data, so perhaps there is a need for explicit feedback mechanisms,
The best way to get feedback is to offer something of value to the user: help in fulfilling their immediate information-seeking task. If a refinement option, “more like this” button or any other mechanism helps users do so, they will pick up on it. Lazy, opportunistic users are our friends!
LikeLike