VISTA Reliability Glossary

The Reliability process for VISTA involves a large number of similar-yet-different mechanisms and structures. This can cause ambiguity and confusion in describing the independent reliability pipelines. For this reason, concepts need to be operationalized and documented clearly to improve discussions. This glossary page will define all key terms across VISTA Primary Outcome Measure (POM) and VISTA Connected Speech (VICS) Reliability.

These are the terms that are related to Reliability and/or steps that feed into Reliability.

Terminology	Definition
Blinding	The process of removing specific bits of information (e.g., patient ID, training condition, observation time point, etc.).
Randomization	The process of selecting a subset from a larger group by chance as a way to reduce potential bias.
Clipping	The process in which recorded videos/audios are segmented by RAs so that the files only include time periods in which patients are speaking.
Clippers	Student RAs who run the videos/audios through the clipping process.
Whisper	This is the step in which audio data is automatically transcribed into a text format which then acts as a foundation for transcribers to work on.
Whisperers/Whisper runners	Student RAs who run audio data through the Whisper pipeline.

Across the projects in the laboratory, there are many types of transcriptions. While transcriptions may look similar across each project, there are key differences described below.

Terminology	Definition
Original Transcriptions	These transcriptions include everything that the patient said. They do not include codes or POM edits. They are saved in the Multiple Baselines data sheet.
POM Transcriptions	These transcriptions include everything that the patient said. They do not include codes but do include POM edits. They are saved in the Multiple Baselines data sheet.
VICS Transcriptions	These transcriptions include everything that the patient said. They do include codes but do not include POM edits. They are saved in the Multiple Baselines data sheet.
CS Transcriptions	These transcriptions include everything that the patient said and are coded for formatting utterance segmentation, etc.
Support Group Transcriptions	These transcriptions include everything that the patient said. They do not include codes or POM edits. These are used for qualitative analysis.
LRT Probe Transcriptions	These transcriptions include everything that the patient said. They do not include codes or POM edits. These are used for naming error analysis.

These are the terms that are often used to distinguish between the different projects and roles that are involved with connected speech.

Terminology	Definition
VICS	VISTA Connected Speech
CS	Connected Speech Battery
VICS Reliability	The Reliability process for VISTA Connected Speech (VICS)
Transcriber	The person who transcribes and codes the audio data into a format that is ready for our linguistic analysis pipeline. The term “Transcriber” is used for VICS, while “Rater” is used for POM.
T1	This is the first Transcriber. This person will receive the uncoded transcript from Rater 1 of POM.
T2	This is the second Transcriber. This person may also act as Rater 2 for POM by having their coded transcript stripped of all coding.

These are the terms that are used to distinguish between the roles that are involved with Primary Outcome Measures.

Terminology	Definition
Rater	The person who transcribes the audio data into a format that is ready for our POM analysis. The term “Rater” is used for POM, while “Transcriber” is used for VICS.
R1	The first Rater is the SLP that completed the session with the patient. This transcription may be used by T1 for VICS Reliability.
R2	The second Rater is second person that will transcribe the audio. This may be the stripped transcription from T2 for VICS Reliability.