Most of the existing methods focus on extracting concepts and identifying the hierarchy of concepts. However, in order to provide the whole view of the domain, the non-taxonomic relationships between concepts are also needed. Most of extracting techniques for non-taxonomic relation only identify concepts and relations in a complete sentence. However, the domain texts may not be properly presented as some sentences in domain text have missing or unsure term of concepts. This paper proposes a technique to overcome the issue of missing concepts in incomplete sentence. The proposed technique is based on the similarity precision for selecting missing concept in incomplete sentence. The approach has been tested with Science corpus. The experiment results were compared with the results that have been evaluated by the domain experts manually. The result shows that the proposed method has increased the relationships of domain texts thus providing better results compared to several existing method.

Non-taxonomic relation is one of the most important components in ontology to describe a domain. Currently, most studies focused on extracting non-taxonomic relationships from text within the scope of single sentence. The predicate between two concepts (i.e. subject and object) that appear in a same sentence is extracted as potential relation. Therefore the number of identified relations is less that what it could be and does not properly represent the domain. In this paper, we introduced a method named Domain-specific Inter-textual non taxonomic extraction (DSINTE) to extract the non-taxonomic relations between two concepts that appear not only in a single sentence but also in different sentences. The proposed method has been illustrated using a collection of domain texts from New York Times website. Recall metrics have been used to evaluate the results of the experiments.