Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2025
|
| Online Access: | https://doi.org/10.5281/zenodo.16418268 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p># Digital Mindfulness Sentiment Analysis Framework</p> <p>This repository provides an implementation of a **sentiment analysis framework** designed to enhance **digital mindfulness interventions** in primary care. It leverages **neural-symbolic modeling**, **hierarchical temporal encoding**, and **adaptive attention inference** to improve emotional well-being and patient engagement.</p> <p>---</p> <p>## Features</p> <p>- **NeuroCognitor Model**: A hybrid neural-symbolic architecture that estimates cognitive engagement states using multimodal inputs (text, voice tone, behavioral patterns).<br>- **Attentive Reinsertion Strategy**: A partially observable Markov decision process (POMDP)-based mechanism that dynamically delivers context-sensitive micro-interventions.<br>- **Multimodal Support**: Integrates text sentiment, interaction patterns, and contextual behavior signals.<br>- **Ethical and Explainable AI**: Prototype-based reasoning and interpretability mechanisms for cognitive dynamics.</p> <p>---</p> <p>## Core Components</p> <p>### 1. **NeuroCognitor**<br>- Uses hierarchical temporal encoding and probabilistic latent-state inference.<br>- Captures **distraction**, **mindful engagement**, and **automatic behaviors**.<br>- Provides attention fidelity scores and distraction indices for dynamic state monitoring.</p> <p>### 2. **Attentive Reinsertion**<br>- A micro-intervention mechanism based on inferred attention dynamics.<br>- Policies optimized to maximize mindfulness while minimizing intrusive interventions.<br>- Ethically adaptive personalization with entropy-based suppression.</p> <p>---</p> <p>## Experimental Results</p> <p>- **Datasets**: Evaluated on UCF-101, HMDB-51, Kinetics, and THUMOS.<br>- **Performance Gains**: <br> - Sentiment classification accuracy improved by **12.7%**.<br> - Patient engagement scores increased by **15.4%** compared to existing benchmarks.<br>- **Ablation Study**: Each component (Multimodal Encoder, Hierarchical State Dynamics, Interpretability-Driven Attention Analysis) demonstrated significant contributions.</p> <p>---</p> <p>## Applications</p> <p>- **Healthcare & Primary Care**: Enhances digital mindfulness apps for patient emotional well-being.<br>- **Behavioral Analytics**: Tracks user attention patterns for wellness interventions.<br>- **Adaptive Interfaces**: Supports integration with mobile and telemedicine platforms.</p> <p>---</p> <p>## Getting Started</p> <p>### Requirements<br>- Python 3.9+<br>- PyTorch >= 1.10<br>- Dependencies:<br> ```bash<br> pip install -r requirements.txt</p>