To Extend or Not to Extend? Enriching a Corpus with Complementary and Related Documents

Abstract

An agent in pursuit of a task may work with a corpus of documents with linked subjective content descriptions. Performing the task of document retrieval for a user or aiming to extend its own corpus, an agent so far relies on similarity measures to identify related documents. However, similarity may not be appropriate if looking for new information or different aspects of the same content. Therefore, this paper combines complementarity- and similarity-based identification of documents, specifically, contributing (i) a formal definition of complementarity using the available subjective content descriptions in the form of relational tuples as well as a taxonomy interrelating the concepts of the tuples, (ii) a technique for classifying complementary and related documents in one go, and (iii) a case study assessing the classification performance for complementary and related documents.