Organizing Knowledge in a Semantic Web for Pathology

Creating Ontologies for Content Representation—The OntoSeed Suite

“…In our own experience in a Semantic Web project in the medical domain (see [22,30] for a longer discussion of this issue, and Section 4.2 below for the project setting), we found that just selecting and extracting relevant sub-ontologies (e.g. from a comprehensive medical ontology like UMLS 1 ) was a very time-consuming process.…”

“…Hence our aim was a more modest, but at the present state of the art of the Semantic Web and in the given application scenario [22,30] a more realistic one: to aid the ontology engineer as far as possible, requiring as little additional resources as possible. Before we come to a description of our approach, however, we briefly review the use of ontologies in NLP, and derive some requirements for "NLP-friendly" ontologies.…”

“…On the other hand, if the resulting ontology is targeted to languagerelated tasks such as semantic annotation, these tasks can be performed more efficiently by means of an ontology which is built in a "linguistics-friendly" manner. On the basis of our previous experiences in applying ontologies to medical information systems [30,22] we identified the following set of operations which can be useful in this context and therefore should be taken into account while conceptualizing the ontology:…”

“…In order to evaluate the costs and the benefits of the OntoSeed approach, we examined two subsequent semi-automatic ontology engineering experiments which aimed at building an ontology for a Semantic Web application in the domain of lung pathology [30,22]. The application operates upon an archive of medical reports (the LungPath-Corpus mentioned above) consisting of both textual and image-based data, which are semantically annotated in order to transform them into a valuable resource for diagnosis and teaching, which can be searched in a fast, content-based manner [22,30].…”

“…The application operates upon an archive of medical reports (the LungPath-Corpus mentioned above) consisting of both textual and image-based data, which are semantically annotated in order to transform them into a valuable resource for diagnosis and teaching, which can be searched in a fast, content-based manner [22,30]. The semantic annotation of the data is realized by linguistically extracting semantic information from medical reports and lists of keywords associated with each of the digital images (both reports and keyword lists are available in textual form).…”