Applying a Reference Collection to Develop a Domain Ontology for Supporting Information Retrieval

Technical documents are often generated during Architecture/Engineering/Construction (A/E/C) projects and research. Information Retrieval (IR) is a common technique to manage growing technical document collections. However, due to the complexity of technical documents, applying IR directly to technical documents often leads to unsatisfactory results. Many semantic approaches, such as ontology, are applied to enhance the performance of IR. Developing domain ontologies requires human effort and thus, is a time-consuming task. Further, developing automated approaches for supporting ontology development has become an important requirement. Reference collections developed to evaluate IR performance can be found in many IR research studies. They are representative subsets from entire document collections and are labeled by domain experts. In order to enhance IR performance, the authors propose an automated approach to grow the base domain ontology from a reference collection. The authors also validate the proposed approach on an earthquake engineering reference collection, called the NCREE collection. The results show that the proposed approach is effective in enhancing IR performance. In addition, the workload of the proposed approach is also affordable for domain experts and can become a possible solution for the automation of ontology development.