Qetch is a tool that allows users to freely sketch patterns on a scale-less canvas to query time series data without specifying query length or amplitude. This project involved developing a novel matching algorithm that accounts for human sketching errors, studying how humans sketch time series patterns and how they preserve visually salient perceptual features while non-uniformly scaling and locally distorting patterns.

This project won Best Paper Award at CHI’18, one of the premier conferences in human-computer interaction, recognizing both the technical innovation and the research contributions to understanding human sketching behavior.

Technical Implementation

The project required building a full-stack application with a sophisticated backend matching algorithm and an intuitive drawing interface. Key technical components included:

• Backend Architecture: Developed in Node.js with PostgreSQL for time series data storage and retrieval, handling complex pattern matching queries efficiently
• Frontend Stack: Built with AngularJS, D3.js for data visualization, Paper.js for the sketch canvas, and Math.js for computational operations
• Pattern Matching Algorithm: Implemented a novel algorithm that accounts for human sketching variations, enabling scale-free and distortion-tolerant matching
• Query Expression System: Designed and implemented support for regular expressions over sketches and relative positioning to query multiple time-aligned series

The system enables construction of complex and expressive queries through two key features: regular expressions over sketches and relative positioning of sketches to query multiple time-aligned series.

Demo Video

Publications