Text Transcription in Labvanced

Overview

Text transcription involves converting spoken language into written text. By transcribing text, the verbal contents become searchable, readable, and analyzable. The primary goal of transcription is to make the content of verbal responses accessible for review, analysis, and documentation.

Examples of Text Transcription in Labvanced

1. Verbal Response Transcription

This use case employs transcription as a core method for data collection. When an experiment requires participants to provide spoken answers, the audio recording object captures their verbal responses. By integrating the relevant events for transcription, the spoken words are then automatically converted into written text. This resulting text can then be saved alongside other quantitative data from the experiment, ready for detailed post-session analysis.

Helpful Demo: https://www.labvanced.com/page/library/75503
In this demo, participants are asked to respond verbally to a series of questions while having the option to preview their responses before submitting. Once submitted, the answer is then transcribed on the screen for a preview of the text transcription feature.

2. Control the Experiment with Voice

Transcription can also be used to create a more interactive and accessible experimental environment by allowing participants to control the experiment with their voice. In this setup, specific spoken words or phrases can be specified to act as commands. When the participant speaks, the audio is transcribed into text in real-time. Labvanced then recognizes this text as a command to trigger a specific action. For example, in a task where a participant must categorize stimuli, they can simply say "True" or "False", or “Left” and “Right”, instead of having to press a key.

3. Real-Time Analysis and Dynamic Task Assignment

This use case takes the ideas from the previous applications and goes a step beyond by using transcription as the basis of real-time, adaptive experimental design. A participant’s verbal response is transcribed and immediately fed into an AI model, like ChatGPT, for instantaneous analysis. This analysis can perform complex tasks such as determining the emotional sentiment of the response, calculating the frequency of certain word types, or classifying the participant’s statement into one of several predefined psychological categories. Based on the output of this AI analysis, the experiment can then dynamically assign the next task, tailoring the experience to the individual's inferred state. For instance, if the AI detects a high level of anxiety in a response, the system could present a calming task next, creating a highly personalized and responsive research protocol.

Transcription Technology

Supporting Integration - Whisper

The transcription automatically occurs in Labvanced via a helpful Whisper integration. Whisper is a pre-trained model used for automatic speech recognition (ASR) that has been trained on 680K hours of labelled data.

Client-side Privacy

As a result of this integration, all processing and transcription occurs solely on the participant’s device. Client-side processing ensures no personal data like voice data is transmitted to third-party servers.