Word-level human interpretable scoring mechanism for novel text detection using Tsetlin Machines

Abstract Recent research in novelty detection focuses mainly on document-level classification, employing deep neural networks (DNN). However, the black-box nature of DNNs makes it difficult to extract an exact explanation why a document is considered novel. In addition, dealing with at word level crucial provide more fine-grained analysis than what available level. this work, we propose Tsetlin Machine (TM)-based architecture for scoring individual words according their contribution novelty. Our approach encodes description novel documents using linguistic patterns captured by TM clauses. We then adapt measure how much contributes making experimental results demonstrate our breaks down into interpretable phrases, successfully measuring