Hakia: Alright I’m impressed.

September 12th, 2007 | by Mike Corey |

hakia logo

So maybe I was wrong. I looked more into hakia today. They actually have a pretty sweet search engine. It seems to utilize wordnet as well as some cool proactive querying. Hakia actually extracts queries from the webpage as its crawled. Thats pretty cool.

Additionally they use a SemanticRank which is allegedly better than popularity rank. This is actually a very useful concept. It seems to be similar to what Chris Anderson talks about in the long tail. While popularity is important in some cases it is generally geared towards the lowest common denominator. For example Amazon.com has information about a lot of things and is very popular. However its safe to say that if I want information about a new phone, a niche blog post dedicated to the phone will most likely contain more information as it is written by subject matter experts in the field they’re interested in.

So how well does it all work? I’m not sure yet. I’ve officially switched my quick search bar from google to hakia for a few weeks (or more if it goes well) I’ll let you (my 5 readers and whoever catches this on technorati) know how it goes.

UPDATE:
Christian from Hakia was kind enough to correct me. Hakia doesn’t actually use wordnet. Its actually ontological semantics. Pretty cool stuff really check out the link in his comment.

  1. One Response to “Hakia: Alright I’m impressed.”

  2. By Christian Hempelmann on Sep 13, 2007 | Reply

    Hi Mike,
    Thanks for checking out hakia! I just wanted to clarify that we’re not using WordNet, which is a common resource in NLP. Its main problem is that despite good coverage, it is very shallow semantically. Very few meaning relations between words are used to structure it, and senses of words aren’t really distinguished (for example, a “table” of five people is a group of humans, while a dining “table” is an artifact made to eat on, usually from wood). The resources we’re using are much richer semantically and our system is aware of such distinctions. You can take a look at our lexicon and the ontology it’s based on at http://labs.hakia.com/OntoSem/hakia-lab-ontox.aspx. More interesting reading on the theory and methodology behind our system can be found at http://www.ontologicalsemantics.com . Maybe you’ll be little more optimistic about AI soon…
    Best,
    Christian
    CSO, hakia.com

Post a Comment