Time2Lang: Bridging AI Models for Health Sensing
A novel approach connecting time series data with language models
This research introduces Time2Lang, a groundbreaking framework that efficiently connects time series foundation models (TFMs) with large language models (LLMs) for improved health sensing applications.
- Eliminates error-prone text prompt conversion by directly connecting sensor data representations to LLMs
- Demonstrates superior performance on mental health classification tasks using wearable and mobile sensing data
- Reduces computational costs while maintaining or improving accuracy compared to traditional methods
- Creates a more accessible pipeline requiring less domain expertise for implementation
This innovation significantly advances digital health monitoring by making AI-driven sensing more reliable and practical for real-world medical applications, potentially expanding access to mental health assessment tools.