The Role of Domain Ontology in Text Mining Applications: The ADDMiner Project

Extracting insights from large text collections is an aspiration of any organization aiming to take advantage of their experience generally documented in textual documents. Textual documents, either digital or not, have been the most common form to register any organization transaction. Free text style is a very easy way to input data since it does not require users any special training. On the other hand, the text material easily collected becomes the major challenge for building automatic deciphering tools. In this paper we present ADDMiner, a text-mining model for extracting causality relationships from a large text collection of accident reports. Our model is based on using domain ontology as well as a corpus-based computational linguistics to guide the mining process. Examples from offshore oil platform accident reports illustrate the potential benefits of our approach.