A Framework for Exploring Computational Models of Novelty in Unstructured Text

Abstract

Novelty modeling in unstructured text data is a research topic within the Natural Language Processing (NLP) Community. Effective novelty models can play a key role in providing relevant and interesting content to the users which is the central goal in many applications including education and recommender systems. This paper presents a framework for comparing different approaches and applications of computational models of novelty in unstructured text data. We focus on computational models that apply methods such as natural language processing and information theory. The framework provides an ontology for computational novelty with respect to the source of text data, methods for representing the data, and models for measuring novelty. We explore the value of the framework by applying it to research on computational novelty in news articles, research publications, books, and recipes. This framework is independent of the type of data in the items and can be used as a tool for researchers to study, compare, and extend existing computational novelty models and applications.