Semantic similarity estimation in the biomedical domain: An ontology-based information-theoretic perspective

Abstract

Semantic similarity estimation is an important component of analysing natural language resources like clinical records. Proper understanding of concept semantics allows for improved use and integration of heterogeneous clinical sources as well as higher information retrieval accuracy. Semantic similarity has been the focus of much research, which has led to the definition of heterogeneous measures using different theoretical principles and knowledge resources in a variety of contexts and application domains. In this paper, we study several of these measures, in addition to other similarity coefficients (not necessarily framed in a semantic context) that may be useful in determining the similarity of sets of terms. In order to make them easier to interpret and improve their applicability and accuracy, we propose a framework grounded in information theory that allows the measures studied to be uniformly redefined. Our framework is based on approximating concept semantics in terms of Information Content (IC). We also propose computing IC in a scalable and efficient manner from the taxonomical knowledge modelled in biomedical ontologies. As a result, new semantic similarity measures expressed in terms of concept Information Content are presented. These measures are evaluated and compared to related works using a benchmark of medical terms and a standard biomedical ontology. We found that an information-theoretical redefinition of well-known semantic measures and similarity coefficients, and an intrinsic estimation of concept IC result in noticeable improvements in their accuracy.

Tracking expertise profiles in community-driven and evolving knowledge curation platforms

“…For example, an assessment of concept alikeness improves the understanding of textual resources and increases the accuracy of knowledge-based applications [73]. The adoption of ontologies during annotation provides a means to compare entities on aspects that would otherwise be difficult to compare.…”

“…Several approaches have been defined for quantifying semantic similarity, the two most prominent ones being: (i) node-based, in which the main data sources are the ontological concepts and their properties; and (ii) edge-based, which uses the edges between the ontological concepts and the edge types as the data source. Note that there are other approaches for comparing terms that don't use semantic similarity; for example, systems that select a group of terms, which best summarise or classify a given subject based on the discrete mathematics of finite partially ordered sets [73].…”

“…These measures, also considered to be hybrid approaches, depend on the empirical tuning of weights according to background ontology and input terms, resulting in ad hoc solutions that cannot be easily generalised [73]. The main advantage of edge-counting measures is their simplicity.…”

“…The first goal is to define a semantic similarity that accurately reflects the semantics of micro-contributions in close proximity, in addition to their degree of specificity in the larger context of the ontology. Hence, a hybrid approach was adopted, which is based on the work of Sanchez et al [73]. More concretely, an edge-based measure (which measures the path between concepts) is redefined in terms of the Information Content (IC) of concepts, which is then expressed via its specificity in the ontology.…”

“…Following the method proposed in [73], the length of the minimum path separating two microcontributions (concepts) -i.e., ( ) is considered. This path evaluates the differential semantic features of both concepts as a function of the amount of non-common ancestors found through the shortest link connecting them.…”