Naikhanov N.V., Dyshenov B.A. —
Computing Semantic Similarity of Concepts Using Wikipedia Links
// Software systems and computational methods. – 2016. – ¹ 3.
– P. 250 - 257.
DOI: 10.7256/2454-0714.2016.3.19560
Read the article
Abstract: The research question is the semantic similarity of terms. The object of the research is the measures of the semantic similarity of terms. The authors consider such aspects as the rationale for the choice of the theme of background knowledge, the construction of a graph of links and measurement of the similarity between concepts. In earlier researches the semantic similarity was calculated based on the statistical characteristics using different contextual analysis methods such as the latent semantic analysis. The given research is the first experience with the reference methods for determining the semantic similarity. Therefore, the focus is made on the ease of calculation steps. Within the framework of the researc determinatino of the semantic similarity is based on the WLM (Wikipedia Link-based Measure) method and similarity measures for separate types of references offered by M. Varlamov and A. Korshunov. In contrast to the well-known measures of the semantic similarity based on the use of Wikipedia database, the measure offered by the authors of the given research uses simple links to Wikipedia articles such as "See also" and "Links" or "External links". This approach allows to improve the performance of the algorithm and is designed for use in applications requiring not a high accuracy of the result but a better performance of the algorithm. These tasks include establishing the correspondence between the competencies of the educational standard and abstracts of curriculum disciplines or the task of analyzing the students' answers to the open questions in the form. The developed measure is cheap, reasonably accurate and accessible.