This is a TTS lab and is the start of the first assignment. Your report for this assignment will contribute 30% of your final grade for the course.

The instructions are here on speech.zone. Start by reading the overview page all the way through, before attempting any practical work. You should note that those instructions assume you are using computers in the PPLS Appleton Tower (AT) labs. Instructions for how to access the AT labs remotely are below.

Next, you need to configure Festival for use in this assignment.

Accessing Festival

You can use the installation of Festival in the Appleton Tower computer labs in-person or using the remote desktop service.

To connect using the remote desktop, follow the instructions here: Module 0 – computing requirements

See the assignment instructions for details on how to download the required voice for this assignment if you have installed festival on your own computer (rather than using the remote desktop to the PPLS AT labs).

Working with the unix shell/terminal

Festival is accessed through the unix shell (i.e., the terminal app). If you’re not really familiar with the unix shell/terminal much before – don’t panic! A good intro to start with is Joe Collins’ Beginner’s Guide to the Bash Terminal (YouTube).  This will show you the very common commands we use in the shell.  The video has topic time stamps so you can go to the demos of specific commands (listed in the info section). You can open up the terminal in linux in the PPLS AT lab (or remote desktop) and try out the same commands as in the video to see what they do.  You may have to reboot the computer and select linux rather than windows (choose the Penguin in the boot  menu).

What this assignment is about

Your mission in this assignment is to discover and explain mistakes made by a typical TTS system. We’re using Festival as our typical TTS system, but don’t focus on Festival as a piece of software, but rather on the processes that are taking place when we convert text to speech.

Do:

• Focus on the theory, as presented in lectures, videos, readings and labs
• Find a wide variety of mistakes and explain these in the context of what you’ve learned about phonetics/phonology
• Provide deep and detailed explanations in terms of the theory, models, techniques, etc

Do not:

• Look for software bugs
• Try to fully understand Festival as a toolkit, read the source code, etc
• Blame Festival for not being perfect: it’s almost always the models and algorithms that make mistakes, not the software.

Please note that the voice we are giving you to work with is not a state of the art voice! It deliberately has a lot of errors for you to find!

Now return to the instructions and start working on the assignment.

Throughout the assignment, you talk to other people in the class to consolidate your understanding of the the theory and to get practical help.

For your report, you should come up with your own (preferably unique) example mistakes and your own explanations of them. You must, of course, write up your report independently and without showing it to any other students.

You can use the speech zone forum on assignment 1 to ask for help.  You might find some of your questions are already answered in previous posts.