README.md

Vashisht2.0

Problem Statement

Develop a deep learning-based and NLP system that detects, classifies, and auto-completes stuttered speech in real-time, improving accessibility and communication.

Team Members:

Raghavan R , Derrick Richard , Tarun Srikumar , Sreenidhi K

Solution Overview:

This ReactJS application leverages deep learning and NLP models to assist individuals with stuttering. It comprises three core modules:

Stutter Detection & Classification:

Utilizes a pre-trained deep learning model (Wave2vec) to analyze audio input and classify different types of stutters (e.g., repetitions, prolongations, and blocks).

Provides real-time feedback to users with visual and audio indicators.

Stutter Correction/Aide with BERT:

Incorporates a fine-tuned BERT model to suggest corrections and alternatives for detected stuttered speech.

Generates context-aware suggestions to guide users in improving speech fluency.

Speech Therapy Component:

Interactive speech therapy exercises with configuration options to modify pacing according to the user's stutter profile.

Running the App:

cd Frontend/FLOWSpeak && npm install && npm run dev

Running the Backend:

cd Backend && pip install -r requirements.txt && python setup.py && python server.py

For the Demo Video Click Below:

For the Setup Click Below:

Real World Applications

Upon implementing Text To Speech, you may implement the stutter correction component to be used in online meetings for people with stuttering (block) problems , to automatically aide them

Practice speaking using speech therapy component before any important speeches for people with stuttering (repetition) problems

We would also love to take inputs from people with ideas on how to help people who have stuttering issues through the contact form.