Ontology-based semantic similarity: A new feature-based approach

SciRecSys: A Recommendation System for Scientific Publication by Discovering Keyword Relationships

“…Vincent Schickel-Zuber and Boi Falting [11] present a novel similarity measure for hierarchical ontologies called Ontology Structure based Similarity (OSS) that allows similarities to be asymmetric. Snchez et al [10] presents an ontology-based method relying on the exploitation of taxonomic features available in an ontology. Markov Chain model which properties are studied in [7,8] is used for computing the similarity and ranking [5,1].…”

“…Sung-Hyuk Cha has conducted a comprehensive survey on probability density functions for similarity measures [3]. Ontology-based methods are also exploited to measure the similarity of keywords [2,11,10]. Bollegala et al [2] use the Wordnet database -ontology of words to measure keywords' relatedness by extracting lexico-syntactic patterns that indicate various aspects of semantic similarity and modifying four popular co-occurrence measures, including Jaccard, Overlap (Simpson), Dice, and Pointwise mutual information (PMI).…”

Semantic based Document Clustering: A Detailed Review

“…As a result, computer understanding of text has acquired great interest in the research community in order to enable a proper exploitation, management, classification or retrieval of textual data [1]. Text document clustering plays an important role in providing intuitive navigation and browsing mechanisms by organizing such large amounts of information into a small number of meaningful clusters [2].…”

Semantic similarity analysis and application in knowledge graphs

“…Because the semantic information they used to define similarity metrics are based on an ontology, those metrics are still applicable to measure similarity in ontology of KGs, such as DBpedia, YAGO, BabelNet, especially for their concept taxonomy. An ontology can be defined as a directed labeled graph, G = (V, E, τ ), where V is a set of nodes, E is a set of edges connecting those nodes; and τ is a function V × V → E that defines all triples in G. Knowledge-based similarity methods measure the similarity between concepts c 1 , c 2 ∈ V , formally sim(c 1 , c 2 ), using semantic information contained in G. In this section, we present the state of the art of knowledgebased similarity methods in three categories according to their properties (Sánchez et al, 2012): (1) based on how close two concepts in the taxonomy are, structure-based methods;…”

“…However, the semantic similarity metrics presented in this paper are used for concepts, rather than words. We convert those concept-to-concept semantic similarity metrics into a word-to-word similarity metrics by taking the maximal similarity score over all the concepts which are the senses of the words (Resnik, 1995;Sánchez et al, 2012). This is based on the intuition that human subjects would pay more attention to word similarities (i.e., most related senses) rather than their differences while rating two non-disambiguated words (Sánchez et al, 2012), which has been demonstrated in psychological studies (Tversky, 1977).…”