SURVEY OF ONTOLOGY LEARNING

Survey of Ontology Learning

Survey of Ontology Learning

1 Introduction

The demand for ontologies has significantly been increased by progress in semantic web research, undertaken to provide common information structure and to solve interoperability problem. Fast and efficient ontologies are a requirement for success of different semantic based systems. Ontology learning aims to reduce cost of building ontology by developing automatic knowledge extraction methods on specified domain and presenting them in an ontology-based structure. As noted by Cruz (2006), ontology evaluation is significant since different structures of ontologies exist that cover multiple domain, which have different sizes, formats, degrees, goals and those that use different representation language. This proposal provides a discussion of the state-of-the-art methods from which ontologies can be constructed from different sources such as structured, semi-structured and unstructured data and knowledge bases and databases with emphasis on the Semantic Web. For each of the state-of-the-art methods, the study will discuss their improvements from its successors.

1.1 Main Idea

The purpose of this research is to provide a survey (overview) of ontology learning techniques.  I plan to discuss different approaches for ontology learning from different sources. The sources that will be examined are structured, unstructured, semi-structured documents, and knowledge bases. For example, I plan to explain approaches for extracting knowledge from an unstructured document (i.e., PDF doc) to form ontology. What is the state-of-the-art approach for doing this? How does this approach improve upon its predecessor? Regarding the improvements, can they be measured in other words; can we define “ontology quality”? Likewise, the same for structured, semi-structured documents, and knowledge bases.

1.2 The objective of the study

To discuss the state-of-the-art methods from which ontology can be constructed given the different types input.

To compare the state-of-the-art methods to their predecessors and explain the differences.

To understand the operation of the structured, semi-structured, unstructured data and knowledge bases and databases on the Semantic Web.

To discuss methods of improvement of each state-of-the-art method. Four types of measures will be applied; this includes class match measure, density measure, semantic similarities, and between the measures.

1.3 Research Question

The proposed study will be based on the following research questions:

Which source provides the greatest challenge of constructing ontology? (For example, is a structured doc more complicated versus an unstructured doc, or does it depend on other factors?)

What approach can be taken to extract knowledge from documents to form ontology?

How can the quality of the learned ontology be measured and improved?

What are the tradeoffs among the popular methods of ontology construction?

2 Literature Review

The Literature Review part will cover the theoretical data related to ontology learning. It will cover the history of the topic along with the application of the different concepts that are related to ontology learning. Therefore, the Literature Review part is going to be an important part of the whole topic.

2.1 Overview of Ontology Learning

The ability to build high quality and practically usable ontologies remains an open research problem. Although there have been increasing improvements in various methods and systems, the ...