Handwriting Recognition. Obviously, handwriting recognition has been mediocre at best on most computers. You have to train the software in your writing style so that it can then translate it into words. Microsoft has many research projects dedicated to overcoming the challenges of handwriting recognition.

Content-based Image Retrieval. Many times, it is impractical to assign keywords to large numbers of images. Microsoft research is working on a project which will take content from the image to figure out what they are, so that they can then assign them to categories and make them searchable. If keywords are available, they will also use them as "cues" to help determine what the image is about. In other words, they will combine keywords (if available) along with image content to determine what the image is about in order to make it indexable and searchable.

Intelligent Video Analysis. Through extraction and analysis of the video, the research department hopes to be able to accurately index video segments and make them searchable. From the research site: "In particular current research focuses on image segmentation, which involves breaking an image up into natural components. This is a useful first step towards scene understanding, compression and representation."

Of course, these are all on the algorithmic side of search and how to make search engines be able to see more content, better. They are also working on a paid advertising option:

Targeted advertising. Using currently available user information (such as demographics, browsing history, etc.), the goal of this research project is to maximize click-through rates to advertising, while still working within the constraints of how advertising is sold. In other words, they will look at what ads the user is particular to, their response rate to the ad etc. to determine what types of ads to serve to a particular user, to help the advertiser reach their quota.

Automatic question answering. Another interesting project, currently called AnswerBot, will retrieve results based on the question asked. Similar to how Ask Jeeves rose to become one of the worlds most popular search engines, AnswerBot will return results which answer the question. For example, a search (on Ask) for the Academy Award winner for best actress in 1962 correctly returns "The 1962 Best Actress award was given to Anne Bancroft in "The Miracle Worker"" from data mined from the IMDB (the Internet Movie Database). Microsoft hopes to be able to duplicate this type of technology.

Microsoft is also trying to develop better data mining and analysis techniques, such as performing a general analysis on a block of data and using that to "fast track" analysis of small parts of that data. For example based on statistical probability, how probable is it that a web page is related to a web site, therefore is complete indexing of the web page needed, or can assumptions be made about the page based on the site. These types of assumptions can help speed up the indexing process.

Overall, I think we have some exciting times coming in terms of search in the coming months and years.

Not only is Microsoft working on these types of indexing algorithms, but you can be certain that the others are as well. We already know that Google has purchased companies like Picasa (to help with image search), and has partnerships with companies like Real Player to improve their multimedia search capabilities.

And obviously they are always working to improve ranking algorithms. Yahoo and Ask/Teoma also have dedicated research departments working on such projects as modeling algorithms and improvements to their paid advertising distribution systems.

There seems to be no limit to the number of PhD's and math wizards working at the major search engines. Google, for example has literally hundreds of employees with PhD's scattered throughout the company. What this also does, however, is make them better and better.

As users get more sophisticated, search engines must grow to meet the challenges brought forth by this new breed of searcher. Most people now are very particular when searching and choosing websites. If they are consistently disappointed and the results displayed they will change search results providers faster than you can say "irrelevant results."

Rob Sullivan
Production Manager
Enquiro.com

Copyright 2004 - Searchengineposition Inc.