In the switch from paper to pixels, the computerized version has typically offered a huge advantage over the original. Spreadsheets, encyclopedias,and communication have been revolutionized by technology. Through all of this, course notes are still generally taken on paper or in a normal word processor. Notes taken electronically currently do not provide many advantages over their paper equivalent.

PixelNotes tries to combine a simple web-based markup language (Textile) with a search engine and Wikipedia-like inter-article links. The inter-article links are a major focus in the software because the ability to have more information on a topic so easily available is very convenient for studying. PixelNotes uses Bayesian classifiers and Support Vector Machines to attempt to automatically identify words for linking. The Wikipedia database dumps have been analyzed to this end.

A big aspect of PixelNotes is determining the applicability of current statistical processing approaches to things like automatic interarticle linking. Enclosing words in brackets is not a labor intensive process, but it may be unnecessary. An automatic method would also enable these annotations in programs outside of the typical wiki environment, providing a convenience for users.

I am collecting data right now on the usability, value, and accuracy of the system. If you find yourself taking notes electronically, consider giving this system a try.