This repository implements a domain-specific, fine-grained topic classification, by training models using textbooks.
Specifically, we aim to construct comprehensive models that accurately reflect the depth and breadth of a particular domain using the knowledge embedded in textbooks. Topic models are used to generate a dataset for training deep learning models to identify the topics present in new pieces of text.
Our motivation is that a dense and comprehensive domain model based on textbooks can serve as a sophisticated framework for educational applications, enabling them to provide enhanced learning experiences via downstream tasks such as learning object classification, automatic index creation, and content personalisation.
Read the associated paper here.
The research is based on the Isaac Alpizar-Chacon and Sergey Sosnovsky's Intelligent Textboks project.