think differently big

Google vs Bing Search

Facts:

1. Google sets up a set of fake results that afterwards show up in Bing results

   1. Each of the fake results returns only 1 result
   2. There’s no mention of how many such fake results have been tested and was the percentage of those showing up in Bing
   3. For some of the fake results, Bing is trying to suggest other searches. Google does not.

2. Microsoft is using clickstream information as one of the ranking algorithm parameters.

   Questions:

   1. Isn’t Google toolbar collecting similar data for the same purpose?
   2. What are the major differences in clickstream collection between the two toolbars?

3. Google accuses Microsoft for “copying” Google’s search results.

My thoughts:

1. If these were fake searches with a single result, there are big chances that the information available through the clickstream was the only available.

2. The fact that Bing tries to suggest other related searches while Google is not doing it makes me think that:

   • not only point 1) above is true
   • but also that Microsoft’s statement of using thousands of parameters is also true (nb: nothing unnexpected here. I think the last search engine using a single set of parameters disappeared 20+ years ago)

My conclusion(s):

1. Google is starting to forget the “Don’t be evil part” and goes after each of its competitors in any way they can imagine (Apple, Microsoft).
2. After leading the search market for 10+ years and owning this market with a very comfortable percentage, Google is starting to feel that competitors are up to something.

Speculation:

The fact that Google accusses such a young competitor makes me think that Google has hit a wall in its attempt to improve its current algorithms and so it starting to fear its competitors.

Disclaimer: This represent a personal perspective on the subject.

What makes a Tweet Relevant?

That’s the question that both Microsoft and Google are trying to answer since they signed deals with Twitter, Facebook, etc. to have these small “snippets of knowledge” included in their search results.

Couple of days ago, Google has announced that they are starting to push out results from Twitter and Facebook (this hasn’t reached all accounts yet though), so you’d expect that they came out with an approach for measuring the relevance. And I have found the following slide (thanks to ☞ TechCrunch):

In case you cannot read it from the low quality picture, here are the 10 metrics, which are pretty cryptic though,

• Language model (?)
• Tweet quality (?)
• Author quality (?)
• Probability of relevance
• Semantics
• Real-time URL resolution (my comment: probably somehow similar to the old PageRank but reversed: the quality of the tweet is determined by the quality of the links included)
• Query registration (?)
• Query hotness
• Query volume fluctuations
• Topicality (my comment: isn’t this more or less the same with the above Semantics?)

Right now, I cannot read between the lines of this algorithm, but there seems to be 3 dimensions that are considered: the author, the tweet and the included links, the query. I guess we will have to wait a bit longer to find out more about it and to see if it works or not.