Enforcing a semantic schema to assess and improve the quality of knowledge resources

摘要

Modern Information and Communication Technologies (ICT) require very accurate and up-to-date knowledge resources, such as databases and knowledge bases, providing information about real-world entities (e.g. locations, persons, events) which can guarantee that results of automatic processing can be trusted enough for decision-making processes. The solutions employed so far to guarantee their quality mainly rely on the automatic application of integrity constraints for databases and consistency checks for knowledge bases. In order to achieve a higher accuracy, there is also a recent trend in complementing automatic with manual checks, via crowdsourcing techniques. This paper presents a methodology and an evaluation framework, based on the definition and application of a semantic schema, which analyses the (sometimes hidden) semantics of the terms in the entity descriptions from a knowledge resource, and allows assessing its quality and the identification of those potentially faulty parts which would benefit from manual checks. The approach is particularly suited for schema-less resources, i.e. resources in which entities do not follow a unique and explicit schema. Our evaluation showed promising results.