Search engine Blekko has been getting a lot of PR from Techcrunch. In short, what Blekko does is to allow users to search the web using tags created by crowdsourcing - read fellow users searching in Blekko. In my humble opinion, Blekko's crowndsourcing method of classifying (adding tags) to sites has limited scope of success. I say this because of two reasons:-
The like and spam buttons are in the SERP (search results page). Can you, as a user, come to a realistic conclusion about the page, as either one that belongs to a particular tag (here in this case - for search query python - the site could about the reptile python or programming language python) ? And is this snippet good enough to know whether the site is a spam? I guess, the answer to both question is a No.
1. The use case for adding tags/classifying a site as spam is unrealistic.
-User searches for a keyword
-Blekko displays the SERP.
Take a look at the anotated picture of the SERP.
Lets continue with the use case -
-User navigates to a site by clicking on a link in SERP.
-User consumes the site content.
-User now wishes to tag the site as "/programming language"; or mark the site as spam
Now what? The only realistic way I can think of is to hit the back button, and get back to blekko SERP, and do the tagging. If the user is already 5-6 pages deep in that site, this is not an option. I couldnt see a easy way to add tags in the current setting. Of course, there are other methods to create slashtags such as adding them through the profile page or using shortcut /add. However, those methods are very cumbersome even for a mild-geek like me.
One possible way out: - It would be nice to explore if having a browser plugin will make adding tags a easier process. For example, if blekko plugin can simply display a lightbox/dialogbox/popup(or, whatever), in each page, and the user simply add tags AFTER reading the site content? Can Blekko build a plugin for chrome browser?
Again, the problem is, this method will require explicit installation at client end, and that requirement will greatly reduce the install base. But the tags are more likely to be relevant than tags added by reading just a snippet in SERP.
The bottom line is, blekko must make adding tags an effortless, and a realistic process, especially for the 2 second attention span Gen Y and Z.
2. Is it realistic to tag the billions of websites by crowdsourcing alone? I think the machine learning approach will be more effective.
Many sites today use Google Analytics for web analytics. (i suspect especially many of these spam sites, and fly by wire content factory sites use GA). For a given search query, I would like to believe that Google might be able to factor in the bounce rates of these pages from the GA data, into their page rank algorithm. I think Google is in a better position to bring in more relevancy by factoring in data from Google Analytics. I am not sure whether Google does this kind of thing, and I think no can ever tell. (Alternative views are welcome.)
PS: Since tags are user generated, I suspect that many variations for the same tag may be found. For example, www.lonelyplanet.come can be tagged as travel, travl, tourism, etc. Blekko would need to have strong tag rewriting algorithms (similar to standard search query rewriting techniques in IR(information retrieval) ) to group similar tags in a topic/genre in intelligent way, and bring more relevancy and precision in search results.
