Slashdot Mirror
Google Delivering Factual Answers
nam37 wrote in about a Macworld article which reads: "Google
Inc. on Thursday began delivering factual answers for some queries at the
top of its results page, to save users from having to navigate over to other
sites and look for the information. For example, if a user enters the query
'Portugal population,' Google returns the answer -- 10.5 million -- along with a
link to the Web page where the information came from, which in this case is the
population page of the U.S. Central Intelligence Agency's Factbook. The
query 'who is Jane Fonda?' triggers the answer '... is an Academy Award winning
American actress, model, writer, producer, activist and philanthropist' and
provides the link to the Wikipedia online encyclopedia's entry for the actress.
A small percentage of queries currently trigger these factual answers, but the
service, called Google Q&A, is in its early stages, said Peter Norvig,
Google's director of search quality."
But, at the same time, they retrieve the rest of your search results. It's not like they tell you they've got the only answer, they just give you what they consider their best answer. Much like, say, a lucky button or something. Only non-optional.
My little site.
I don't know that "factual" is a good term for a lot of the stuff on Wikipedia, especially "contested" articles that tend to go through revert wars and lots of vandalism.
Google may be cool, but most of its algorithms and technology are closed. We have no idea how accurate the information will end up being, and also, how corruptible.
How do you trust information from Google?
The same way you trust ALL information you find on the Internet. YOU DON'T!
WELCOME TO THE INTERNET!
Basic research skills
Do not trust one source of information - always corroborate it with another source.
If one website says that the population of Portugal is 10.5 Million and another one says 20.5 Million, then there is obviously an error somewhere. If the second one says 10.1 Million, then you could probably live with the difference.
Of course, how many 'average users' trust everything they read on the internet blindly and would never think to question the information?
This sig is intentionally blank
And its a different system than the one that copes with "speed of light". Its missing stuff like "price of oil" or "price of gold". It gets "area of missouri" but not "area of germany". It seems to be triggered on a small set of keywords and an associated set.
If I were them I would negotiate with AFP, Reuters so that the indexing Robot obeyed a delay time, since even slightly stale news, say 15m for FOREX and Equity prices makes the information unusable for trading.
But, very good, keep it up Google, and show M$ what real innovation is about.
Also, how does it determine which sites are authoritative in this manner? I'd say it's 99% likely they'll go by PageRank. That's how they rank 'authority' for everything else - relevancy, accuracy, etc., in the end. Given 2 pages with the exact same content, the number of incoming backlinks (IBL's) will determine which site G. chooses - and the PR is a composite of IBL's + PR of pages they're on.
"This is your life, and it's ending one minute at a time."
how can they determine facts?
If a user asks: who is George Bush for example, is the right answer:
A. The current president, (blah blah, blah)
or
B. A Moron who (blah, blah, blah)
The point being that the "facts" are sometimes in the eye of the beholder.
Case in point, the CIA may state a different population than the country itself believes.
Looks like google is the one playing catch up to microsoft this time. Microsoft search has had this feature since it was in beta. And it even gets teh president in 1996 question correct.
+ united+states+in+1996&FORM=QBHP
http://search.msn.com/results.aspx?q=president+of
no, those are actual responses... i tested it myself.
Post-rock/Ambient/Drone and other noise.
erm, actually it -is- a fact that he is the current president of the USA, but it's purely subjective that he's a moron.
Having read the PageRank paper, which is apparently the backbone of their search engine technology, I'd have to say that they openned at least part of their technology to peer review.
That said, as far as question answering is concerned. Question answering systems are an active area of Natural Language Processing research. If you are curious about them, you can easily get your hands on a paper or two on the topic by Googling "Question Answering Systems."
How is google supposed to survey the upteen billion pages, and decide what is subjective and what are the facts?
Well, opinions are statements about your state of mind while facts are statements about the world. Calling someone a moron is an opinion because it reflects your state of mind regarding that person and calling someone the holder of the office of president is a fact about the world. So you'd need a lexicon of opinion oriented words and then entries could be parsed for loaded language like "moron" in the "who is" category for example.
The difference, naturally, is that Microsoft uses only Encarta for its results, whereas Google, at least in theory, uses the entire web to parse its results. (In practice of course, most of the results seem to be coming from either Wikipedia or CIA's factbook, but still)
More than mere navel gazing.