Ravi’s Blog

I got some feedback about my recent post about vertical search.

1. Discovery and extraction of vertical entities is extremely hard. I agree. However, this is a hard problem in the general context. If it is a random webpage and we are trying to categorize it and extract an entity from it, then this is a fair statement. However, the advantage of vertical search is that we apply the knowledge of a particular vertical. For example, instead of crawling all the web and trying to categorize sections of documents as a job, we seed the crawler with some sites. We tune the extractor to certain pages etc, simplifying the problem space a great deal.
2. It’s all about relevance. Someone pointed me to this post from mark pincus of tribe.net. He seems to echo the same thoughts as me, in that, relevance is the key. Also, i didn’t know that walt mossberg wrote about vertical search as well.