According to an article appeared on PCWorld earlier today, the Indian Yahoo! team would be developing an innovative search technology that, in response to a search query, will offer users summaries of web pages rather than the usual list of links and text tokens from the page itself.

After the "Glue Pages" experiment, a service in beta stage that offers results combined with images and other media all in the same page, the Yahoo! team is now looking for yet a new paradigm in its attempt to try and regain search engine market share over Google.

Rajeev Rastogi, vice president of Yahoo Labs Bangalore, said in a recent interview that while search engines typically list a series of URLs with their search results, sometimes — such as in Google's case — featuring a short snippet of the page content, this isn't often enough for the end user, who still has to try and guess which is the correct search result.

Yahoo's approach is to offer meaningful content summaries rather than disconnected chunks of text. The goal, it is apparent, is going to be quite hard to reach: in all likelihood, the team will need to develop complex natural language processing AI algorithms that will be able to parse the content (be it text, images, Flash, or other media), understand its overall meaning and finally return an informative summary along with the site URL in response to an appropriate user query.

Similar algorithms are currently being used for contextual advertising, a field in which Google's superiority is still undiscussed: Yahoo's estimated figures of potential yearly earning that could have come from a Google-Yahoo advertising deal is a testimony to that superiority.

Ap

parently, the search engine is therefore starting with an initial handicap, although Rastogi seemed confident of the team's possibilities: the feature is due to go live on the Yahoo! homepage next year, while the staff is currently at work developing efficient data mining algorithms to extract relevant information from the Web.

Rastogi also told PCWorld that the format of the results will vary based on the type of the page: product pages will feature pictures, name of the manifacturer and price, while an abstract of — for instance — a hotel-related page would contain contact information as well as a map with directions.

Google already offers similar features, although part of the site abstracts is coming directly from website sitemaps, files in XML format uploaded by the webmaster that helps the Mountain View, Calif.-based search engine parse the content and tells the Google spiders exactly which links should be featured in the abstract. Yahoo's approach does however involve a completely automatically generated summary, which could be a daunting task to implement correctly.

Search results customization seems to have become the latest trend among the "big two". Earlier this year, Yahoo! released SearchMonkey, a platform that allows developers and site owners to customize the presentation of URLs in search results; more recently, Google added a feature that Google Account owners can use to personalize the order in which results are being displayed, also enabling users to add notes next to each result.

But Yahoo's Bangalore, India team is working on other search refinements as well: a search suggestion and autocompletion tool is in the works and will be available next year, adding category suggestion as well. According to Rastogi, the new technology will prompt users to help them refine their queries so that they can narrow down the category of URLs that have the information they are actually looking for, based on the user's search history.