MSc Computer Science Final Project - University of Birmingham 2019

  Improving Accessibility: Telephone-based Automatic Speech Recognition for Users with Disfluent Speech
                                  
                                   Project Grade: Distinction
                                Degree Classification: Distinction

This repository contains all of the files submitted as the final project for an MSc in Computer Science at the University of Birmingham.

The MSc primarily taught programming through the use of the Java programming language. However, due to the web-based nature of the project and the wide availability of open-source speech recognition modules, I decided to use the Python programming language in conjunction with HTML and CSS. A small amount of acquired (i.e not authored) JavaScript was also used.

The report (included in repository) charts a research project which identifies a set of accessibility problems generated by ASR systems when processing disfluent speech and/or are used by disfluent users; specifically, it will focus upon those users who stutter/stammer. It elaborates upon the problem domain in order to settle upon a suitable focus before moving on to the iterative design, implementation and testing process behind the creation of a series of prototypes designed to investigate potential solutions. It concludes with a ‘final’ system proposal which proports to solve the identified problems before providing an evaluation as to what worked and what didn’t throughout the duration of the project.

Methodology:

The project is composed of several web-based components:

1. A Data Collection Application - records speech samples and stores data in MySQL database.

   -> Used to collect speech samples for use within the project

   -> One prototype developed - Web_Application_Data_Collection_Prototype

   Note: you are free to download the prototype but the database connection has been terminated so it will throw an error if you attempt to record any data.

   -> One deployed application - Web_Application_Data_Collection_Deployed

   Note: the deployed data collection application was originally deployed on Google's App Engine platform. Since the project has concluded this has been taken offline. You are free to download the final application but functionality might not match that described in the report.