litewi: A combined term extraction and entity linking method for eliciting educational ontologies from textbooks

• Published in: Journal of the Association for Information Science and Technology Vol. 67; no. 2; pp. 380 - 399
• Main Authors: Conde, Angel, Larrañaga, Mikel, Arruarte, Ana, Elorriaga, Jon A., Roth, Dan
• Format: Journal Article
• Language: English
• Published: Blackwell Publishing Ltd 01.02.2016
• Subjects: Construction; Education; Extraction; knowledge acquisition; Knowledge representation; Languages; Learning; natural language processing; Terminology; text mining; Textbooks
• ISSN: 2330-1635; 2330-1643
• DOI: 10.1002/asi.23398

Major efforts have been conducted on ontology learning, that is, semiautomatic processes for the construction of domain ontologies from diverse sources of information. In the past few years, a research trend has focused on the construction of educational ontologies, that is, ontologies to be used for educational purposes. The identification of the terminology is crucial to build ontologies. Term extraction techniques allow the identification of the domain‐related terms from electronic resources. This paper presents LiTeWi, a novel method that combines current unsupervised term extraction approaches for creating educational ontologies for technology supported learning systems from electronic textbooks. LiTeWi uses Wikipedia as an additional information source. Wikipedia contains more than 30 million articles covering the terminology of nearly every domain in 288 languages, which makes it an appropriate generic corpus for term extraction. Furthermore, given that its content is available in several languages, it promotes both domain and language independence. LiTeWi is aimed at being used by teachers, who usually develop their didactic material from textbooks. To evaluate its performance, LiTeWi was tuned up using a textbook on object oriented programming and then tested with two textbooks of different domains—astronomy and molecular biology.