13 Mar. 2007
Search 2.0 - What's Next?

NLP (Natural Language Processing) powered Powerset

While still in stealth mode, Powerset has already raised $12.5M in pre-money valuation from several venture capital companies and angel investors like Reid Hoffman, Luke Nosek and early Googlers Aydin Senkut and Zain Khan. The difference between Powerset and the traditional search engines is that while typical search engines like Google and Yahoo don't take into account stopwords (by, after, the, etc), stopwords are a very important part of the engine for Powerset. Why? Because Powerset relies on a semantic capability that can be triggered by using these stopwords. So while the "book by children" and "book for children" queries return exactly the same results in Google, Powerset evaluates them separately and somehow cares about your stopwords as well.

Personalized Search

Palo Alto based Collarity is a very new company entering into the personalized search area. The question that pushed them into this challenge is: "Why are your search results exactly the same as the next person's search results?" This is not a very new idea - Google (with its Kaltix acquisition in 2003) and others already offer this feature, albeit weakly. However Collarity seems very strong with their innovative interface (Collarity Slider), outsourced approach (Collarity Compass) and promising technology.

Social Search

Read/WriteWeb has covered the area of social search very thoroughly already in two articles in July by Ebrahim Ezzy. Two good examples are Eurekster's Swicki and Rollyo. Swicki is a community-driven search engine that allows users to create deep, focused searches on a specific niche - and 'learns' from its community. Rollyo allows users to create and publish their own personal search engines, based on websites they decide to include in their "SearchRoll".

Image Search

Image Search has been around for a very long time, but to be frank it's still very primitive. What most image search engines do is just look for text around images and examine the image tags.

Riya was the first to introduce advanced face recognition technologies in image search. This obviously requires a lot of computing power and just because of this, Riya's weekly burnrate is supposedly over $100K. Co-founded by web 1.0 veteran Munjal Shah and face recognition gurus Burak Gokturk and Azhar Khan, Riya is now entering a whole new space - "search by likeness" with like.com. This may come in very handy, for example when you try to find a watch that is similar to the one you have a digital photo of. That's why Riya is expected to make partnership deals with, or get acquired by, e-commerce companies like Amazon and eBay. It's worth noting that Riya was once in acquisition negotiations with Google, but this never happened - and Google ended up acquiring another face recognition company, Neven Vision. So we can conclude that Google is pursuing this technology very closely!

Approach Enhancements (Vertical Search)

Vertical search is a relatively new discipline in search. Basically, vertical engines look up a very limited subset of the internet - so they are more efficient than generic search engines. Because their search area is not so broad, they can adapt themselves for the specific needs and common points of their area of focus. We won't go in too much detail about vertical search engines, as it has already been covered in a recent article in Read/WriteWeb. But we can categorize the major vertical engines this way:

• Jobs: SimplyHired.com Indeed.com, Bixee.com (India), Eluta.ca (Canada), Recruit.net (Hong Kong)
• Travel: Sidestep.com, Kayak.com, Mobissimo.com
• Health: Amniota.com, CloserLookSearch.com, GenieKnows.com, Healia.com, Healthline.com, Kosmix.com, MammaHealth.com, Google Health
• Classifieds: Edgeio.com, Oodle.com
• Blogs: Technorati, Bloglines, Blogger Search, Sphere, Feedster
• Source Code: Koders.com, Krugle, Google Code

Conclusion

The innovation in search does not stop and there's much to look forward to in the search space. What's more, Google and Yahoo search APIs and the open source Nutch and DMOZ projects allow anyone to try out new ideas. Nutch, supported by Yahoo and shielded under Apache Software Foundation, is providing a free global search engine. DMOZ gives you a very large open source web directory edited by volunteers.

Google will have a hard time competing not only its big adversaries like Microsoft, Yahoo and Ask - but also the ambitious startups that are opening new dimensions and bringing forth new approaches. We will probably hear of acquisitions in this space as well.

We may not have covered all the promising new search offerings here, so please let us know your feedback in the comments below. Also let us know which of the above approaches sounds the most promising to you - and why.

Previous