Evolution of Semantic Annotations

SCHEME: CORE

CALL: 2014

DOMAIN: IS - Business Service Design

FIRST NAME: Cedric

LAST NAME: Pruski

INDUSTRY PARTNERSHIP / PPP: No

INDUSTRY / PPP PARTNER:

HOST INSTITUTION: LIST

KEYWORDS: Ontology evolution, semantic annotation, medical ontologies, annotation quality

START: 2015-09-01

END: 2018-08-31

WEBSITE: https://www.list.lu/

Submitted Abstract

The use of Knowledge Organizing Systems (KOS) like ontologies in the medical field has show great possibilities to tackle semantic interoperability issues. In many cases, KOS elements serve to annotate information in order to make their semantics explicit for machines, which facilitate the automatic treatment, and in particular, the retrieval of the annotated information. However, the medical domain is highly dynamic by nature. Since 50% of the knowledge of this field is renewed every 10 years, the content of KOS, representing this knowledge, has to smoothly follow the evolution of the domain to remain exploitable by underlying software applications. As a consequence, the modifications made in these KOS, based on domain evolution, must also be propagated to all elements that depend on the KOS. This is the case for semantic annotations.In this project we will propose a formal framework for the (semi-)automatic maintenance of semantic annotations. The proposed approach will consist in (i) understanding the quality of annotations through manual and automatic annotation processes (ii) identifying and characterizing the evolution occurring in KOS and in (iii) exploiting this information to define algorithms to maintain semantic annotations affected by KOS evolution. Moreover, two maintenance scenario will be distinguished: (i) direct migration of the annotation in the case where those are modifiable and (ii) an ad-hoc migration process in the case where annotations are not modifiable. In the latter, the goal it to ensure that annotated data are still searchable. We will provide a query enrichment mechanism that will integrate evolution information at query level to retrieve relevant information. The proposed approach will be validated on real case studies consisting in the maintenance of semantic annotations attached to Electronic health records contained in the Luxembourgish eSanté platform.

This site uses cookies. By continuing to use this site, you agree to the use of cookies for analytics purposes. Find out more in our Privacy Statement