Article Preview
TopIntroduction
An ontology is a formal, explicit specification of a shared conceptualization. The ontology should be machine readable, with explicitly defined types of concepts and constraints on their use are. It should capture consensual knowledge, that is, it is not private to some individual, but accepted by a group.
Ontologies are often created and updated with human intervention. They represent reality, and therefore require frequent updates. Both the creation and the update of ontologies are costly activities. To overcome this problem the discipline of Ontology Learning has emerged. In this paper we refer specifically to Ontology Learning from text.
Surveys conducted since the early days of Ontology Learning show the different methods used to tackle the problem. The majority of methods follow an approach named the Ontology Learning Layer Cake. This approach splits the task into multiple steps towards learning an ontology, namely term extraction, concept formation, creation of a taxonomy of concepts, relation extraction and finally rules extraction.
This strategy does not take into account the conditions of the problem, nor the quality of the results obtained using the Ontology Learning Layer Cake methods. Working with well-formed text (i.e., books, edited journal and other quality sources) is not the same as working with potentially lower quality sources, such as emails.
Moreover, splitting a task into smaller, sequential tasks, often help reduce complexity without undermining the results. As this paper shows, splitting the tasks, as in Ontology Learning Layer Cake methods, results in low quality results, making the methods unviable.
The paper is organized as follows:
- •
This introduction
- •
A review of the literature regarding Ontology Learning from text
- •
The target of Ontology Learning
- •
The input used for the Ontology Learning task
- •
The Ontology Learning Layer Cake approach
- •
Alternatives approaches
- •
Discussion and conclusion