Ontologies are widely used in information systems. Major problems in building ontologies include a lack of standards to integrate, the use of fixed categories based on a single viewpoint, and the absence of automatic knowledge acquisition methods.

The authors of this paper propose an automatic ontology building approach that will employ only a simple, manually developed kernel as its basic structure. The kernel has the essential meta knowledge to manipulate the ontology elements.

To test and implement this approach, the authors’ group introduces a Hasti system that uses a natural language (Persian) as its input, and outputs a learned lexicon and ontology. The authors go into great detail in their discussion of the system architecture and components; the learning algorithms were also elaborated clearly throughout the paper. The evaluation of the Hasti system was based on small-scale test cases in a wide range of domains. The process was separated into two phases to obtain a precise evaluation, and various values of system parameters were applied to test system performance in different conditions. The experiment was comprehensive, and the results provide a fair view of every aspect of the system.

The automatic learning of ontologies is clearly the solution to ontology creation. The approach proposed here, and the ongoing Hasti project, provide a new direction and approach to solving existing problems in ontology building.

The paper is very well organized from a scientific viewpoint. As is noted in the conclusion, more work needs to be done, including trying to adapt the system to other languages, and further testing and evaluation of the system.