Slashdot Mirror
Microsoft To Launch 'Question' Site
prostoalex writes "Microsoft will try to make the search process more social, Business Week reports, by creating a question-and-answer Web site. They certainly are entering a quite crowded niche." From the article: "It's one of the many ways that Web companies, including Yahoo and Google, are trying to set themselves apart with social search, a targeted pursuit of information that's influenced by the preferences of a person's peer group. Social search is a method whose time has come, Osmer says. Microsoft research shows that generic search engines can't answer 50% of queries asked, he says. The new tool, whose name he didn't disclose, will be 'one of the larger projects for us' this year, Osmer says."
There are so many questions to ask about this...my first response is that attempts to make information more friendly don't seem to have that great of a track record. Does anyone remember "Ask Jeeves"? Compare how its interface competed with the super-minimalist interface of google.
Anyway, there is this one quote:
Microsoft research shows that generic search engines can't answer 50% of queries asked, he says.
What type of questions were they asking it? Were they factual questions, like "What is the Capital of Burundi?", or were they process oriented questions, such as "How can you make cookies that are not too hard, but are cooked all the way through?" The first question, if you type "capital Burundi" into google, you get an answer for. Trying to search for information on the second would be much harder, I imagine.
Hopefully I didn't put any [] around my words.
> targeted pursuit of information that's influenced by the preferences of a person's peer group.
Excuse me, but when I am looking for information, finding only what my peers think is good for me is the last thing I would want. Social conformity is the death of truth.
What if you could do a honest search that did not factor in the prior popularity of the site, but relied on other criteria, so that a new site with unique content might have a chance of getting found?
Oh, you mean, so my search results would begin with nineteen pages of splogs? Bright idea. I wonder why no-one's thought of it before.
I'm not wrong. You haven't thought about it hard enough.
I can see you've never worked a library reference desk. Sadly, half the time the person asking the question doesn't understand the terms they're using to ask. Rule number one in reference work is never to trust what the patron gives you.
... and thought they could drive there)
I've had people come in asking:
Where are your car books? (they wanted a bio of Mario Andretti)
"How do you burn stuff?" (they wanted info on pyrography)
"What do you have on crafts?" (they wanted to know how to carve pumpkins)
"I need a map of the world." (they wanted to plan a trip to Egypt
"Where is that ambulance book?" (they wanted the World Almanac)
"What can you tell me about Greece?" (they wanted the price for a 1943 coin from there)
Now, being a librarian, I can ask clarifying questions and figure out more precisely what they're looking for. Thus far, search engines have proved to be very very bad at doing this. If Microsoft's upcoming site proves to be better at it, more power to it. But all the hype about AI notwithstanding, computers have a very long way to go to be able to do it half so efficiently or perceptively as a human being.
Putting the burden on the seeker to "reformulate the question" probably works well for most Slashdotters (given that they tend to show above average intelligence and articulation), but assumes far too much intelligence on the part of the average seeker.
How can a post be modded "overrated" or "underrated" when it hasn't been rated yet?
It's 42. The hard part is knowing the question.