Voice Emotion-AI

Tara

Has been created to help humans take care of their mental well-being.


Functionalities

Tara is a Voice-AI Agent that allows our users to interact with WeVoice program, she has been trained by experts in Cognitive Behavioral Therapy (CBT), Neuro-Linguistic Programming (NLP) and Rapid Personal Transformational therapy (RPT).
Tara is capable of providing meaningful conversations and executing all the functions of the transformative coaching program. The more you interact with her, more effective she can be to you,  recommending effectively, and providing valuable and personalized plan alongside providing  valuable and actionable insights to reflect on your life-patterns.

The most important functions of Tara are the following:


Monitoring: Tara analyzes all input data to show your mental state and progress during the program.

Emotion Recognition: The voice emotion recognition module analyzes you to identify relevant patterns in your mental state changes. 

Mental Wellbeing program support: Tara guides you during the sessions of the program, allowing you to access all the functionalities and services of WeVoice.

Education & Transformation: Tara supports your awakening and transformation of patterns that don’t serve you.



Architecture

Through transfer, reinforcement and deep learning techniques, Tara continuously improves to communicate more deeply; she is available to understand you and guide you through all the sound healing programs. Tara is an integration of six deep learning modules

Automatic speech recognition (ASR): This module transforms your voice into text using an accurate ASR system, it guarantees Tara can clearly hear you.

Intent detection: Tara is trained to understand your intentions using transformer models and transfer learning techniques.

Voice emotion recognition: Tara listens to you to perform speech and spoken discourse analysis to evaluate your mental state.

Language generation: The conversations are generated from two deep learning modules, a text generation model based on GPT3 that produces relevant content according to the conversation and a question answering model to generate appropriate questions and respond accurately during the program.

Dialogue manager: This component is the brain of Tara, it includes the system logic, connecting all the above modules in a unique integration, to support the users correctly in each step of the transformational program.

Text to speech synthesizer (TTS): All the text information produced by Tara is transformed into speech. This module allows Tara to speak using natural voice sound.


Properties

Tara works in a very demanding environment, where it is needed to combine many different types of data and information. In order to produce high-quality integration, Tara’s models have the following properties:

Multimodal integration: Tara analyzes your mental state and emotional map to guide you during the program, offering the most suitable therapies, communicating with empathy according to your current state.

Contextual data:  All the models are trained by experts, providing Tara with knowledge on NLP, CBT and RPT therapies.

Continuous learning: Tara keeps learning about you, improving text generation, intent detection and question answering models through reinforcement learning techniques.