Intelligent Filtering of Electronic Messages

Dr. Fredrik Kilander, K2LAB

With the ever increasing pressure from providers of electronic information the possibility of automatic filtering becomes attractive. The term filtering is usually associated with weeding which discards the unwanted, but it can just as easily be regarded as sorting in order of interestingness. Filtering is also separate from information retrieval in which a query is aimed at an information store; filtering regards a continuous message flow as being less predictiable wrt content and the end user is commonly without premeditated preferences.

The IntFilter project at DSV has focussed on the categorization of messages from Usenet News with examples provided by the end user as the primary source of sorting. Other systems rely on different means of discrimination and they will be reviewed along with the workings of PEFNA, the working prototype system used in IntFilter.