Text segmentation is the task of dividing a document of text into coherent and semantically meaningful segments which are contiguous. This task is important for other Natural Language Processing (NLP) applications like summarization, context understanding, and question-answering. The goal of this project is to successfully implement a text segmentation algorithm. We take a supervised learning approach to text segmentation and propose a neural model for this task. We aim to extend this task to podcasts by using existing transcription services. Our model obtained a Pk score (described below) of 6.54 on the Wiki-50 dataset, which was an improvement over our baseline score of 69.23. We experimented with self-attention as a modification to our model.
