English sentence unit detection based on text features

Abstract

Sentence unit detection in automated speech recognition (ASR) system is crucial for enriching the ASR output, improving the human readability, processing the word stream of the output and bridging the gap between the ASR system with the NLP applications.This thesis presents the machine learning models for sentence unit detection in written text. In this context, sentence unit (SUs) is referred to punctuation marks in the sentence, in which the focus is on adding period (“.”) to the unstructured word sequence.