Introduction to the special issue on machine learning methods for text and images

The amount of unstructured and semi-structured data available in the form of text documents, images, audio, and video files by far exceeds the volume of data stored in relational databases. Yet, in order to optimally utilize this data, it is necessary to devise methods and tools that extract relevant information and support efficient, content-oriented access to it. There is considerable interest, commercial as well as academic, in this domain which has sparked research in machine learning methods for information access. Some of the devised techniques have already led to improved tools that are incorporated in today's information retrieval and knowledge management systems. This special issue arose from a NIPS 2001 workshop on Machine Learning Methods for Text and Images held at the Whistler Resort, Vancouver, Canada. The aim of the workshop was to present new perspectives and new directions in information extraction from structured and semi-structured data for machine learning. Contributions to the special issue were also open to researchers who had not presented their work at the workshop, and 18 papers were submitted. After rigorous reviewing, only 5 were accepted---an acceptance rate of less than 30% for this special issue. The original workshop was jointly sponsored by KerMIT 1, an EU funded project investigating the use of kernel based methods in the analysis of text and images and NSF-ITR/IM 2 an NSF funded project on statistical learning technologies for digital information management search.