Ravlt Audio Gui

In this folder, you'll find a

1. environment.txt or environment.yml (should be interchangeable for our purposes) which lists all the dependencies needed for this project. You can make a new conda environment using:

conda env create -f environment.yaml

or simply create an empty environment and import the dependencies listed in req.txt manually

conda create --name my_environment

2. the main file called ravlt_audio_gui.py. You can run this file by typing the following command into the terminal

python ravlt_audio_gui.py

USAGE

3. After launching the code, a dialog box prompting the user to input their initials. This is just to keep of what member's making what edits.
4. After pressing 'Submit', a GUI will appear.
5. Getting Started: Press Import to import a partcipant folder listing the audio transcriptions. Once imported, you'll see 2 other windows populate: the Top Left displays the .csv as edits are made; the Top Right displays the waveform for the audio (this will be necessary for making Edits and Inserts (see below).
6. Once imported, the name of the block (e.g. Block 1 Recall) will apppear in the Left Column. Once ready to begin, press Start and the audio will begin playing. Upon pressing Start, the current word will appear at the bottom of the Middle Column as well as be highlighted in the Top Left window.
7. Quality Check: Bottom Right of GUI Once the audio plays, the first thing we want to check if the speech was correctly transcribed. You do this by pressing Accept to validate the transcription, For Review for audio that may be inaudible and needs an additional listening, and Off Task for any speech that was not an attempt to recall a word from the list (e.g. 'I'm tired', 'My parent shoud be coming to pick me up soon'). Additionally, you can press Add Note to add any information about a decision related to the quality check or edits.
8. Important Note: For an utterance to qualify as Off Task, it must not relate at all to the task. Any repeats, mistakes (e.g. says 'drape' rather than 'curtain'), subvocalizations (i.e. quietly repeating words to themselves as a memory strategy), or otherwise do not count. However, if a participant says a word from the list (e.g. 'parent') out of the context of a recall (e.g. 'My parent shoud be coming to pick me up soon'), this should be labeled as Off Task.
9. Edits: Middle of GUI There are 3 types of edits that are able to be made here: Edit, Drop, or Insert. To initiate an Edit or Insert, press Start Edit/Insert. Once the start button has been pressed, you can then provide Time Data and/or Word Data. To Edit the onset/offset of the current row or provide the Insert with the onset/offset, click on the waveform, first, where the onset will be and, second, where the offset will be. For Edits, the red highlighted area over the current word will be updated; for Inserts, a blue highlighted area will be shown on the waveform. Additionally, to Edit the word at the current row or Insert a word, press Save Edit/Insert and a text box will appear. Once you press Continue, changes will be saved (you can press Refresh on the Left Column to see your updated changes). Finally, to Drop a row, simply press the Drop button.
10. After finishing all checks on the word, you will press Next Row to move on to the next word.
11. Additional Usability: There's a Search button at the Bottom-Right of GUI that, once pressed, a text box can be used to search for a specific word or onset (e.g. if the .csv lists 'Curtain' being said at 6.4s, you can type in 'curtain' or 6 and press Search). Otherwise, the Restart, Refresh, Play Full Audio, and Zoom buttons are pretty self explanatory.
12. Progressing Once finished with a row, press Next Row to continue to check the next one. Once finished with a file, text will appear in the Bottom-Right of GUI saying 'Press the 'Start' button the begin the next file'. Continue validating until all files in the folder have been validated.

Next Steps:

1. Fix it so you can make an edit and an insert at the same utterance
2. Change it so you can skip to the next file before pressing the start button so you don't have to toggle through every row.