MSN Search Engine - history and highlights

The MSN Search engine is fighting for dominance with the Google Search Engine and Yahoo Search Engine.

In early 2005 MSN stopped piggy-backing off Yahoo and introduced its own search engine. Like the other major search engines, MSN employs its own pure, algorithm- driven technology.

MSN search engine: differences

The MSN Search Engine includes some features that differ from those found in Google, Yahoo and Ask Jeeves. For instance, there are options beside the search box which allows you to perform focussed searches. Some of these searches are unique to MSN. 

MSN allows search in news, local and image indexes. But perhaps the neatest, and most unique, feature is the ability to search the Microsoft Encarta encyclopedia.

Also, beside the search box, there is an option to use a "search builder". This offer the same ability as "advanced search" features in other search engines. Some users might find it easier to use. It's worth a look, you might find it works better for your needs.

MSN search engine: ranking pages

How does the MSN search engine rank pages? the answer is, in much the same way as other search engines. It uses text in content, anchor text in linksheading tags, alt tags, the title attribute in links—anything visible to the user in a standard browser—to weight its search results. But, like most other search engines, it does not use the meta keywords tag as a weight—it's not visible and has a history of being spammed. The tags <b> vs. <strong> and <i> vs. <em> all have an equal weight.

MSN use  a subset of 569 different properties to predict the relevancy of a particular document. This is more than most other companies, so you may find that pages which take account of any conceivable property do particularly well in the MSN search engine.

Microsoft research labs are avidly pursuing search technologies, and might be going in some interestingly different directions. For instance, Microsoft added MSN RankNet, a neural network, to its search engine algorithm in June 2005. It learns from human search patterns.

Andy Edmonds, lead program manager, and Erik Selberg are two of the geeks who work on making MSN Search better. They had a frank talk about how the MSN Search engine works on Channel 9 Forums in October 2005. They said what they're doing to beat the competition. 

Some points they make:

To be really effective, the "neural net" needs to parse the writing on the page in the same way as a human being. (I''l let pass the question of: "Which human being?) The neural net needs to "understand" the concepts being presented from the site. Google, basically, just ranks pages based on how many people reference them.  But this can be exploited, as with the "miserable failure" bug.  Parsing the thought on the page is hard, and (although they're not saying) its unlikely that MSN have this any where near cracked.

The technology

Super high end 64 windows machines using parallel processing of neural nets.

Advertising

Google hasn't nailed advertising. The long tail has not been dealt with. Finding obscure products for small numbers of people is an MSN priority.

For details of topics like aggregation and query expansion, Erik has a lot of information on his site. He created dogpile, the metacrawler InfoSpace is based on.