While looking around for some tokenization databases to help me build a term weight analysis tool, I came across a Microsoft article from January of 1997 on the subject of text analysis. This paper discusses some of the elements used in text processing, topic detection and classification and clearly shows that in 1997, search engines weren't used keyword density anymore than they are today. Since there had been some debate on this subject, I started a thread on the subject over at SEOChat.