Would anyone know or have any details what some of the early employees are doing? would be an interesting list.

The “20% time” is a great idea, but everybody’s pet project can’t actually be worthwhile. Just my $0.02 (anyone got change for a nickel?)

The “20% time” is a great idea, but everybody’s pet project can’t actually be worthwhile. Just my $0.02 (anyone got change for a nickel?)

Steve Robertson and Karen Sparck Jones, two grandparents of the search industry (they have been working on search since the 1960s), wrote a tech report in 1997 entitled “Simple, proven approaches to text retrieval”. In this paper, they outlined some of the most basic things one can do to get better results. These are techniques and approaches that have proven, time and again, over the years, to work.

One of the first things they mention is iterative searching, i.e. giving the users the ability to do relevance feedback in order to improve term weighting. After 8+ years of daily search improvements, has Google offered this capability yet? No.

Another thing they mention is longer queries. Does the user want better results? Then they have to use more words to specify what they want. So it would seem to me that Google would want to come up with better interfaces for eliciting more information from the user. While the average length of a user query has increased slightly over the past 5 years (from like 1.8 words to 2.2 words or something like that), users are still too lazy to type in a 10-word query themselves, it’s true. But if the search interface itself provided some way of easily allowing the user to increase the size (and therefore specificity) of their query, results would also improve.

So where is that interface? Why does the SERP page continue to be crowded with ads, instead of iterative search and query specification aids? These are things that have been known for going on 30 years now to improve search. And yet Google continues not to offer them.

Why?

I know I’ve asked this question on this forum, and on many other forums, over and over again. The only response I have ever heard from Google are statements along the lines of: “We here continue to evaluate numerous clustering, query modification, etc. algorithms, but have so far not yet found one that meets our quality standards.”

Well, that was believable for a while, but now that it has been 8+ years, that explanation is starting to wear thin. How can it be that 30 years of scientists find that longer queries and relevance feedback can improve the search process, but Google can’t find a single way of implementing this?

Maybe it is the case that Google has found that these long-proven search improvement techniques are actually no longer proven, when you have web-scale data collections. (Frankly, I think it would be just the opposite; more data means better feedback models.) But maybe Google has found this to be true. If so, then publish, guys, publish! Give back to the community!

There was a thread about this on Battelle’s blog a month or two ago. If you’ve really found that certain things no longer work, publish, and give back to the community. Because as it stands right now, from the outside, it looks like you actually haven’t been testing any of these ideas (or at least only half-heartedly), and instead have been wasting your time developing instant messaging clients and finance applications, instead of organizing the world’s information by focusing on search.

I can’t speak to every area at Google, but I certainly wouldn’t agree with that as a sweeping generalization. In search quality, for example, lots of people come in every day and brainstorm ways to make search even better.