In this entry in MSN Search's WebLog, Ken Moss, the MSN Search General Manager, explains that a new ranking technology is being employed by MSN Search. An artificial neural net is now being used as the basis of the MSN Search ranking algorithm. For those interested in additional details of Microsoft's vision of this technology, read the patent abstract (but be warned, the language used in the abstract is not for the layperson).

A little research on the topic will quickly convince you that a neural net is not something the average person can comprehend easily. Artificial neural nets are models inspired by the study of biological neural networks, and consist of a collection of interconnected units that perform a collective task. The units are modelled using mathematical functions, which themselves depend on the other functions in ways defined by the structure of the network. As you can imagine, the larger the network, the more complex the set of interrelations between units can become.

One of the fascinating, and very desirable features of neural nets is their ability to learn. This is undoubtedly the reason that MSN (and very likely other major search engines) have chosen to make use of this type of model. The goal of these search engines is to provide the most relevant search results possible for a given search term. The neural net model produces an improved set of results over time, as it learns from past searches. Ken Moss illustrates this concept with the graphic displayed in his blog entry. In his example, the relevance of the search term "pbs evolution videos" has improved over a period of a few weeks because the MSN search engine "learned" to produce more relevant results.

All of this information leads us to the question of what exactly are the factors that the algorithm considers relevant? The answer is exactly what any search engine optimization (SEO) professional already knows: relevant content, good web site design, and inbound links. These are the fundamental principles behind SEO that we have been emphasizing here on marketposition.com for years.

Creating relevant, high quality content is critical to the success of your web site. First identify the keywords that are most appicable to the theme of your web site. Next, write compelling copy that makes use of the keyword(s) throughout the page. Keep in mind that it's best to focus on a single keyword or phrase on each page. If you have identified several keywords, then you will need to create several pages to accomodate them. Use a search engine optimization tool like WebPosition's Page Critic to help you optimize each page for the keywords of interest.

The design of your web site can have a profound affect on your search engine visibility as well. There are a few important concepts to consider when working on your web site design. First, a consistent web site theme should be developed. Web site content that revolves around a common, central theme generally performs better in search engine results than web sites containing mixed content. Secondly, create a solid network of links throughout your web site so that a search engine spider (and your visitors) will have no trouble navigating around the web site. Create a site map that contains links to all pages on your web site that contain content you want to ensure is indexed by the search engines. Link the site map to your home page to ensure that a search engine spider can find it easily. Be sure to make use of the robots.txt file on your web server to help direct the behavior of search engine spiders that visit your web site. Finally, use an HTML validator (like this one provided by W3C) to validate your HTML code. The validator will help you identify any poorly formatted HTML code on your pages. This is important because improperly formatted HTML code can cause problems for search engines when they are attempting to index your content.

Search engines measure the number (and quality) of links pointing to your web site from other web sites. It's therefore very important to work on increasing the number of incoming links from other quality web sites, preferably sites in the same general content theme as your own. Read this article for a good overview of things to consider when working on Link Popularity.

Conclusion:

The intelligence of neural nets is great news for the search engine user, who will find relevant search results more readily. It's also good news for SEO professionals because it will reward efforts to conform to the search engines' terms of service. Those who choose to violate the engines' terms of service, however, will not be happy with these developments because efforts to artifically inflate search engine rankings are more likely to be discovered as the neural nets "learn" over time. For a review of illicit practices to avoid, read this article from the marketposition.com archives.

                 ← What is this?

---