Entity-Event Ontology Construction by Conceptualization of Mentions in Text Corpus

Abstract

Knowledge-based systems constitute a powerful tool for tackling and navigating complex domains, but they have the potential to be employed more often in practical tasks if some obstacles are cleared. Creating and keeping knowledge bases up-to-date is a challenging problem without automatic extraction of knowledge from data sources like documents. One of the solutions is ontology learning, which enables automatic construction and population of ontologies used to store knowledge. This chapter proposes an automatic method for domain ontology construction based on extracting entities and events from texts. Also, it is stated that upper-level template ontologies used when analyzing text corpus are suitable for creating target instance ontologies that describe a specific domain. The task of instance ontology construction is formulated in the terms of reconstructing real-world events via analyzing their mentions in a text corpus and structuring them according to the template ontology. This method allows an automatic analysis of big volumes of textual data like posts from social networks, news, contracts, specifications, etc., by utilizing natural language understanding tools used to extract domain knowledge. We developed a system that collects texts from the Internet, analyzes them, builds an ontology, and presents it as a knowledge base. One of the current applications is in optimizing business processes in a domain of civil aviation: document management, sorting and navigating documents, text summarization, semantic enterprise search, and exploratory search. Furthermore, it is claimed that extracted knowledge can be used to construct informative features in machine learning tasks.