Towards Semantic Quality Control of Automatic Subject Indexing

Abstract

Automatic subject indexing is a key technology for digital libraries, however, factors like concept drift hinder its success in practice. Releasing high-quality results into productive retrieval systems may still be possible when thorough quality control is applied, which may support algorithmic improvements and allow to create high precision filters. Since errors and their relevance can depend on characteristics of concepts and their relations, evaluations should take semantic aspects into account. For this reason, we present the prototype of a web-based reviewing tool which especially aims at fostering semantic analysis and visualization, that is, considering relations, properties and semantic categories of concepts, algorithms and reviews. The tool uses techniques of the Semantic Web. Its application is demonstrated by example.

Fusion architectures for automatic subject indexing under concept drift

“…Section 4.2), which has not been indexed by humans with STW concepts before. Detailed quality judgements have been collected using a web-based tool [33]. A group of 6 human indexers rated samples at the document level on a 3-point scale (reject, fair, good) and at the 22 In STWFSA, we added special processing routines.…”