Techniques for Automatic Metadata Assignment and Evaluation

Jason Scheirer, Software Developer, University of California, Riverside, iVia/DataFountains Project

Jason Scheirer graduated from the University of California, Riverside in 2005 with a degree in Linguistics and Computer Science. He researches computer simulations of linguistic phenomena and statistical natural language processing. He currently works for the iVia/Infomine project on improving classification accuracy in DataFountains.

Abstract

This presentation outlines the techniques described in the paper Developing Practical Automatic Metadata Assignment and Evaluation Tools for Internet Resources (Paynter 2005). The automatic assignment of metadata to internet resources is a large problem of immediate practical use to managers of large digital libraries and metadata collections. The generation of metadata is a complicated process that includes different evaluation and generation criteria for several different classes of data to be generated. The iVia Virtual Library Software Project approaches this by creating an evaluation workflow using incremental, hybridized human evaluation and objective statistical evaluation tools rather than a one-time evaluation of the generation tools. Repository maintainers using the automatic tools can assist in improving the machine's results and increase confidence in the results of automatic metadata generation and extraction tools.

Slides