Tuesday, February 03, 2009

Idea #17 - Search is about keyword

Have you had the experience of going through the overly crowded results from Google search, only to realized that the content that you are interested was buried in the 7th "o" page, or you need to do multiple searches with different combination of keywords just to make yourself understood. Exactly. The Web is filled with good information, but computers do not, yet, understand the semantics, or meaning of their relationships.

Semantic Web is trying to solve it all by a new way of designing web sites, which is ambitious, but I don't know how it will turn out. TechCrunch review on this.

I propose keyword-keyword search, i.e. keyword in, keyword out.

For example, if you type in "George Bush", you get a page of other keywords such as, "president", "white house", "Texas governor", "George H. W. Bush", "Walker", "Dick Cheney", "Republican", "War on terror", etc. a ranking set of keywords that are all associated with George Bush in one way or another. Each will also be a hyper link that explains why these two terms are associated.

There is an internal database that stores the prioritized relationships between any pair of keywords, phrases, or concepts, which is learned by mining billions of web pages, news articles, documents, and books publicly available.

Think about it. Isn't this how our brain remembers and learns new things ?

Note, a linguistic approach was taken by PowerSet, and a number of other approaches are being worked on by companies.

Update 1: application in brand advertising. Nielsen Online creates Brand Association Map (BAM) that gives a detailed analysis of the competitive brands, product categories, key attributes, market segments associated with a branch, which I believe are derived from data mining the massive articles from the websphere and blogsphere. This can be very useful for a marketer to create a targeted strategy. Imagine this association map applied to any person, event, concept, or object!

Update 2: Ervi has a similar goal of making sense of the web by connecting contents associated with each other.

No comments:

Post a Comment