Article type
Year
Abstract
Background: Answering complex biomedical questions is possible if the knowledge encoded in documents is well identified and represented in a way so that machines can process it and find implicit relations across documents and over the larger Web of Data (WoD). Ontologies enable information to be inferred and entities identified; they make published information machine-processable and enable data interoperability.
Objectives:
We want to:
1. build our vocabulary;
2. define a participatory methodology for maintaining and developing the ontology;
3. set up the collaborative software infrastructure supporting the methodology; and,
4. establish the governance structure.
Methods: We are adapting previously proposed methodologies. We have analysed topic lists from Cochrane groups, examined our corpus against existing biomedical ontologies, defined use cases, identified terminologies that represent our corpus, and brought them together into a single vocabulary. We are studying various ontology governance structures, we are also evaluating tools to facilitate the participation of a decentralized community, e.g. Cochrane Information Specialists, in the evolution and governance of the ontology.
Results: The first version of our ontology makes it possible to identify and instantiate the PICO (population, intervention, comparison and outcomes) model. Our ontology brings together terms, metadata and properties from SNOMED, ATC, RxNorm and MedDRA. We have many thousands of terms properties and instances. Also, we have used our ontology for annotating approximately 300 documents.
Conclusions: Developing ontologies in the biomedical domain is a multidisciplinary-participatory exercise. Although there are various efforts building and maintaining ontologies in a collaborative decentralized fashion, tools and methodologies are not yet fully mature as to be readily applicable to various scenarios. Representing knowledge in Evidence Based Medicine is an emerging field, one in which Cochrane is playing a leading role.
Objectives:
We want to:
1. build our vocabulary;
2. define a participatory methodology for maintaining and developing the ontology;
3. set up the collaborative software infrastructure supporting the methodology; and,
4. establish the governance structure.
Methods: We are adapting previously proposed methodologies. We have analysed topic lists from Cochrane groups, examined our corpus against existing biomedical ontologies, defined use cases, identified terminologies that represent our corpus, and brought them together into a single vocabulary. We are studying various ontology governance structures, we are also evaluating tools to facilitate the participation of a decentralized community, e.g. Cochrane Information Specialists, in the evolution and governance of the ontology.
Results: The first version of our ontology makes it possible to identify and instantiate the PICO (population, intervention, comparison and outcomes) model. Our ontology brings together terms, metadata and properties from SNOMED, ATC, RxNorm and MedDRA. We have many thousands of terms properties and instances. Also, we have used our ontology for annotating approximately 300 documents.
Conclusions: Developing ontologies in the biomedical domain is a multidisciplinary-participatory exercise. Although there are various efforts building and maintaining ontologies in a collaborative decentralized fashion, tools and methodologies are not yet fully mature as to be readily applicable to various scenarios. Representing knowledge in Evidence Based Medicine is an emerging field, one in which Cochrane is playing a leading role.