An unofficial blog that watches Google's attempts to move your operating system online since 2005. Not affiliated with Google.

Send your tips to gostips@gmail.com.

December 15, 2009

Google Real-Time Irrelevance

Google decided to show real-time search results for some popular queries, but the implementation failed miserably. Google forgot to focus on relevancy and started to show irrelevant results from Twitter. Just because someone posts uninteresting messages about a popular topic doesn't mean that the tweets are suddenly relevant.






An article from Wired quoted Google's Amit Singhal, who explained that "Google Real Time search is Google's relevance technology meeting the real-time web":
Google said its real-time search offers not just a stream of data, but an organized stream filtered of spam and other irrelevant information. The key to keeping the updates relevant is that Google judges "author quality," "probability of relevance" and "query hotness," according to Singhal. "That's what real-time search is all about," he said.

From what I've seen so far, Google's real-time search shows the limitations of Google's relevance technology. Until these limitations are addressed, Google should stop cluttering the search results with irrelevant Twitter messages.

16 comments:

  1. Agreed. But what's funny, i don't see the Twitter results anymore.. Have they read this post? =]

    ReplyDelete
  2. Irrelevent results could be due to the topics themselves being irrelevent !!

    ReplyDelete
  3. I'm confused... What about those search results isn't relevant? They seem right on to me.

    ReplyDelete
  4. I would have to disagree. I mean you are getting your information from microblogging sources... do you expect dissertations on your search query? I recently searched for the new legend of zelda movie just released and got the link to the movie on daily motion from some random guy's twitter. everything, whether what i was looking for or not, that scrolled by related to my search.

    Someone searching for tiger on the day of his scandal is getting the majority of the results they would have the day before. except for on realtime results.

    Context for the tiger search result:

    1. news (relevant)
    2. images (irrelevant, actual tigers)
    3. tiger woods official site (relevant, but prolly not what you were looking for)
    4. tigerdirect (irrelevant)
    5. us census bureau (wtf?)
    6. realtime results (all of which relate to the tiger woods scandal, most likely what you want when searching for tiger today)

    I found many more interesting links and comments on realtime search than i did going through the usual results.

    ReplyDelete
  5. Considering that almost every post on Twitter can categorically be labeled as pointless, irrelevant spam it is not a big shocker to see that stuff in the feed.

    What the feed offers is some insight into the mental workings of our population, which is down right scary, as I am sure most of us already know. That being said, it is a frightful reminder of the lack of IQ in the country and I am not sure I want to be reminded every time I search.

    ReplyDelete
  6. The biggest advantage of Twitter for Google is NOT the content to index, it's LINKS. Yes, because thanks to Twitter and its 140 characters limit you have to choose very accurate words to describe links, and Google loves that to know what pages are about.

    ReplyDelete
  7. @Jacob Trust me it isn't any particular country that has the lack of IQ. That is a universal truth.

    ReplyDelete
  8. Beyond the relevance, what bothered me personally was the fact that it kept pushing down results before i got the chance to check them out. Design wise i think the feature as it was implemented comes a bit in contradiction with the auto-generated style of Google News. Since it's the real time thing we're interested in more than the actual '140 characters' content why not limit it to one shown tweet at any given time. Maybe that can be put in a box which can be expanded or collapsed by the user...

    ReplyDelete
  9. It's not much of a clutter. It's just a little box.

    ReplyDelete
  10. I enjoyed the article and the discussion. I believe Google rushed this feature out without adequate scrutiny. I also believe that there is lot of value in real-time data such as tweets. TipTop is my preferred search tool for such data. I also do all my online shopping there.

    ReplyDelete
  11. Couldn't see the point of your aricle Alex. Agree completely with MiniV.

    ReplyDelete
  12. So if you search for "dexter", would you like to see results like "dude, have you seen the latest Dexter? WTF"? Do you think that chat replies like "@rougeroses Totally missing the point mate #ratm" are relevant results for [RATM]? Google shouldn't display random Twitter messages just because they happen to be recent.

    "In the context of information science and information retrieval, relevance denotes how well a retrieved set of documents (or a single document) meets the information need of the user." [Wikipedia]

    ReplyDelete
  13. Well, if you're looking for recent news on "dexter", then "dude, have you seen the latest Dexter? WTF?" tells you that the latest episode of Dexter is surprising/exciting/confusing. That's relevant to the query.

    "Google shouldn't display random Twitter messages just because they happen to be recent"? Doesn't "real-time search" convey an emphasis as much, if not more, on "recent" as on "relevant"? (You could even say that for real-time search recency IS relevancy.)

    You want the most relevant AND recent information, sure, but it sounds like what's going on is that different users place different importance on recent vs relevant qualities, which is certainly understandable, but not necessarily Google's fault here.

    I suppose Google may eventually desire to tailor their real-time results to each user's taste, with a customized balance of recency vs relevancy. Or better decide when its appropriate to show real-time results at all for a certain user and query.

    In its current iteration, I think Google's real-time results already offer an interesting and useful glimpse of the latest zeitgeist.

    ReplyDelete
  14. Couldn't agree more with the article. All RTS does is clutter the search results with irrelevancies, make the page load slower, and force me to pause an animation before I can concentrate on the actual search results. Ridiculous.

    If I wanted search results that have nothing to do with what I queried, I'd use Yahoo or Bing. Or, for that matter, Twitter itself.

    ReplyDelete
  15. Could not agree more with Prolorn.

    ReplyDelete
  16. I COMPLETELY AGREE as well.

    Searches for realtime type of contents are very unique and specific in its application and I would rather Google have kept it as a seperate category. I think this is an example of Google’s universal search on mad.

    I have blogged about some specific scenarios for searching through real time contents here where I would just go directly to Twitter search.

    http://www.search1x.com/2009/09/20/what-treasures-can-you-find-on-twitter/

    ReplyDelete

Note: Only a member of this blog may post a comment.