Tutorial

This is a “tutorial” designed to make working with the aligner easier for people. In it, I'll mention how to build dictionaries while using praat, how to assign phones to the words that will be in these dictionaries, and how to use them all together in the aligner.

In terms of building dictionaries, there wasn't much issue with that so long as you are familiar with programs like praat and audacity. It was quite straightforward as you would just transcribe the audio within praat, which lead to no problems. To save time you can use the praat script. Afterwards, take the audio, and run them in praat along with the praat script(syllable extractor?) to assign silences and sounds, making your transcription life much easier. If the audio is not .wav convert them to that format somehow, such as exporting them to .wav in Audacity, like I did. Once you have transcribed your audio, without punctuation of any kind and in all caps, you are free to to then move on, to make the dictionary.

Assigning phones was also not too difficult a process, the only hiccups occurred once again due to a lack of fundamental understanding of things like vim. Watching this video helps http://www.youtube.com/watch?v=J1_CfIb-3X4 . Knowing the commands in vim, such as global find and replace(:%s/what you want to find/what you want to replace it with/g) as well as the commands for things such as line breaks(\n), or to include things that appear on the line before or after what you're searching for (^.*) makes this process of trimming the unnecessary information from the text grid quite easy. Thankfully knowing these command are simple and the only real work needed in this step was to keep a careful eye out to not bleed your results by using global find and replace. For example, you can take a word like ox and put it into its phone components by making it (I think this is right) AH K S. Spacing between the segments representing the phones, as well as capitalisation of them is necessary. If it is necessary for you to edit and isolate instances of “I”'s from the text you've obtained from your transcription, make sure that later on when you use find and replace on other words, you do not accidentally find and replace the characters needed to represent the phones from what you have previously done. This problem can be solved by either using find and replace in a way such that they are ordered and neat, making bleeding impossible, or you can just make everything you find specific enough to avoid this(eg :%s/ I / IY /g rather than :%s/I/ IY /g). Either way, being careful about what you use find and replace on is a necessary evil.

Things to improve on in terms of better understanding the aligner would be to know more about the details of how exactly it works. This eventually comes in handy when it's time to fine tune the aligner. This includes being more well versed in things such as what exactly pruning and pruning numbers are, as well as other components within the aligner. Being aware of that, as well as knowing that capitalisation, as well as a lack of punctuation in the text is necessary for the aligner would make things much easier.

Useful commands:

(the process to add things to git hub)

git status -s

git add filename

git status -s

git commit -m “fixes #XX”

git pull origin master

git push origin master