Gold standard based evaluation of ontology learning techniques

Abstract

A growing attention has been paid to the ontology learning domain. This is due to its importance for overcoming the limits of manual ontology building. Thus, ontology evaluation becomes crucial and very much-needed in order to select the best performing ontology learning method. The aim of the present paper is to offer a new method for assessing a learned ontology in comparison to a gold standard one. In order to avoid issues of previous precision and recall measures, the proposed method is based on a new ontology disambiguation engine. The latter provides meaning annotations to concepts. Next, we propose a set of measures that exploits the meanings of concepts to evaluate the learned ontologies. To prove the efficiency of the proposed solution, we conduct a set of experiments that test our method on well-known ontologies. Experiments show that these measures scale gradually in the closed interval of [0; 1] as learned ontologies deviate increasingly from the gold standard.

An effective automated ontology construction based on the agriculture domain

“…Words that come under different labels are given more importance when compared with the words that fall under a single label. The extracted vocabulary obtained is then processed by extracting the composite terms with a length threshold of three. This length is additionally validated by verifying the AGROVOC. The POS tagging serves as a linguistic filter to extract complex terms. The learned ontologies are evaluated using the gold standard [31] for the layer of lexical terms. The quantitative description of the dataset is presented in Table 8. A sample data were given to the system, as shown below.…”

An ontology based modeling framework for design of educational technologies

“…There are several ways of evaluating ontologies in the literature such as gold standardbased, corpus-based, task-based and criteria based (Brank, Grobelnik, & Mladenic, 2005;Cristani & Cuel, 2005;Sfar, Chaibi, Bouzeghoub, & Ghezala, 2016). However, these approaches are for evaluating ontologies, and not directly for an ontology framework, which is the case in this paper.…”

A survey of ontology learning techniques and applications

“…At first, this technique utilizes a denotational mapper known as ‘lexeme-to-concept’ to map the extracted ontology. Finally, semantic similarity is computed through WordNet using various measures: Leacock and Chodorow ( 89 ), Zavitsanos et al ( 90 ), Trokanas et al ( 91 ) and Sfar et al ( 92 ) assess the learned ontology by comparing it with a gold standard ontology . The proposed approach computes the similarity of two ontologies at lexical and relational level by transforming the ontological concepts and their attributes into vector representation.…”