Biomedical concept taxonomies or, more generally, ontologies are abundant and they provide concept inventories that can be used in semantic indices. So, we use the concepts instead of keywords for the questions.
Hierarchical classification algorithms (Silla and Freitas 2011) can classify documents and questions onto the concepts of these inventories, facilitating the matching os, and also structured data (e.g., RDF triples) that already have explicit semantics based on the same concepts.
Classify documents
(1)read the questions, which could not be submitted to the current bibliographic databases directly.
For example, for this question,Q1,’What is the role of thyroid hormones administration in the treatment of heart failure?’
(2)translate the questions to collections of terms(in this case, it can be called concepts, using MESH) For example, for the question above, we might get T1, “heart failure thyroid hormone treatment” crossponding to it, and every term or phrase can be seen as a concept, which should be stored in the cas.=> using the specified concepts as the MESH filters
(3)T1 then should be submitted to a document retrieval engine as queries, and the engine returns a (possibly long) list of documents.
(4) Studying these documents to find snippets that contain information answering the questions
Pipeline
Biomedical concept taxonomies or, more generally, ontologies are abundant and they provide concept inventories that can be used in semantic indices. So, we use the concepts instead of keywords for the questions.
Hierarchical classification algorithms (Silla and Freitas 2011) can classify documents and questions onto the concepts of these inventories, facilitating the matching os, and also structured data (e.g., RDF triples) that already have explicit semantics based on the same concepts.
Classify documents
(1)read the questions, which could not be submitted to the current bibliographic databases directly. For example, for this question,Q1,’What is the role of thyroid hormones administration in the treatment of heart failure?’
(2)translate the questions to collections of terms(in this case, it can be called concepts, using MESH) For example, for the question above, we might get T1, “heart failure thyroid hormone treatment” crossponding to it, and every term or phrase can be seen as a concept, which should be stored in the cas.=> using the specified concepts as the MESH filters
(3)T1 then should be submitted to a document retrieval engine as queries, and the engine returns a (possibly long) list of documents.
(4) Studying these documents to find snippets that contain information answering the questions