User-Centered Maintenance of Concept Hierarchies

1 Introduction

Beside the full-fledged ontologies that are supposed to form the basis of the Semantic Web, other constructs exist to represent background knowledge. Many of them have a long tradition and history that goes back well before the birth of the Internet, including glossaries, dictionaries, vocabularies, gazetteers, taxonomies and thesauri, just to name a few. They all, including ontologies, have in common that they are used to represent and organize knowledge in a structured way to fulfill specific functions for different purposes, thus we refer to all of them as Knowledge Organization Systems (KOS). A subset of them have in common that they organize the concepts representing single units of knowledge in a hierarchical way, which we call concept hierarchies.

Concept hierarchies have numerous important applications from improved indexing of document collections to faceted browsing and semantic search applications. But the creation and maintenance of concept hierarchies is cumbersome and very time consuming. On the other hand, many concept hierarchies already exist and more and more of them become publicly available, ideally as linked open data, and can be reused for different purposes. If an existing concept hierarchy is to be reused, several tasks have to be performed, reaching from the proper selection of the source to start with to the adaptation for the desired purpose which includes deletion of unnecessary concepts, merging and splitting of concepts and especially the addition of missing concepts.

In this chapter, we describe several successful approaches to the semi-automatic creation and maintenance of different types of concept hierarchies. These approaches have in common that they incorporate the human expert as a central part of the system.