Automatic topic detection for spoken input

In the Natural Language Understanding field, one of the important tasks is topic detection. Given the recent advancements in the field of deep learning, we explore the task of automatic topic detection for spoken input in the news domain. In this work, we study the effect of applying deep learning techniques such as Recurrent Neural Networks and Word Embeddings to the task of topic segmentation and topic detection given a spoken text stream. We also study the performance of the system given both written and spoken text using our novel approach for text segmentation.