Raju vegesna’s Blog

I have been to MacWorld today and was surprised to find the no of add-on modules for Apple’s succesful iPod. I guess this is the story with any succesful product as there will be n number of other add-on products for every succesful product.

One common add-on I found though is an iPod car holder which sits on the charger slot. For this add-on, these vendors had few cool cars to display this add-on. But by design, the user must use his head sets.

Can someone tell these vendors that Californian drivers cannot use headsets while driving according to califronian law (which makes their add-ons useless - and ofcourse when the show is in SFO)?

Raju

I have seen google yahoo and other search engines talking about a common problem - understanding the context of the user’s query. Basically, if a person searches for ‘Java’ for example, the challenge is whether to return coffee related results or Java programming language related results. There could be n number of meanings to a search phrase and understanding the context will be a challenge for search engine creators.

Thinking about this problem on my drive to the office today, I feel that this can be partially addressed when search engines can understand the user’s previous search queries. A general user could end up searching with different keywords to find the right results. If the search engine can save the user’s recent queries (last day/week/month etc) on the local machine, the search engine can sort of get an idea of the context of user’s search to some extent. Based on this information, the search engine can refine its results to that context.

Another approach is to get a context on user’s work type and behaviour by understanding his work based on the documents on this machine etc. Once a machine is searched and indexed, the search engine can obviously get some idea on the type of work that individual does and it can then fine tune the web search results based on the context of his work. This will not work in all cases though.

Now having mentioned these two approaches, the challenge is on the technical aspect of implementing these. To address the first problem of saving all user’s recent search results, there has to be a local service on the machine tracking all of these results.

To address the second problem, there has to be a local program running that can index all the user’s files.

Both of these can be addressed by providing a desktop search which will index the local files and will also observe user’s recent search activity. Once this information is available, providing context to a search should be relatively easy. I am not saying that it’ll completely address the issue, but atleast can provide better results to certain extent.

This can further be improved when other information is also taken into consideration like indexing your emails (gmail, yahoo etc). This way, search results can further be optimized.

I am hoping that Google and others have this behind their plan of releasing a desktop search. I’ll wait to see how they execute.

Raju