ASK: acquisition of semantic knowledge

• Authors: Trevor P. Martin
• Venue: ICANN/ICONIP'03 Proceedings of the 2003 joint international conference on Artificial neural networks and neural information processing
• Year: 2003

Any computerised information storage system contains assumptions about the form and content of stored information, and the nature of queries. Most obviously, retrieving data from a relational database assumes knowledge of tables and attribute domains. In semi-structured and unstructured data, assumptions may be less explicit but are still present. For example, using a TF-IDF index assumes that the user is aware of the "correct" keywords to be used in queries. One way around this is to implement an ontology, i.e. a "concept dictionary" indicating sets of query terms which are equivalent and containing a hierarchy of concepts e.g. plant is a supertype of tree, which in turn is a supertype of oak. Such a hierarchy can be used to generalise or specialise queries. Manually creating an ontology is a very labour-intensive process. In this paper we describe a system which automatically acquires a concept dictionary. The concept dictionary should be regarded as a property of the whole system, i.e. the data and the querying mechanism, not just the data. It makes term similarity explicit and can form the basis for personalisation, by automatically translating a user's terms into those understood by the system.