Date of Graduation

Summer 2024

Degree

Master of Science in Computer Science

Department

Computer Science

Committee Chair

Razib Iqbal

Abstract

Voice-enabled interactions have become increasingly popular with the rise of voice assistants. Identifying contexts or meanings from voice commands and conversations with smart assistants can contribute to the autonomous control of smart home devices and appliances. To improve automation, there is a growing need for efficient context detection that eliminates the need to memorize voice commands. To address this need, I followed a two-step approach in my research. In the first step, I developed a unique context recognition model using a transformer, an attention mechanism, and a fully connected neural network. I trained this model on a conversational dataset of daily interactions with a smart assistant, which demonstrated higher accuracy than some baseline models. In the second step, I developed a dynamic context detection framework that groups new conversational data to automatically identify new contexts and merge these newly identified contexts with the initial context recognition model, which helps the model learn new contexts and adapt to a changing environment. After carefully analyzing the collected data, I conclude that this two-step approach enables dynamic context detection in smart homes while interacting with smart assistants to automate various day-to-day actions.

Keywords

clustering, IoT, natural language processing, pre-trained model, transformer, word embedding, zero-shot learning.

Subject Categories

Artificial Intelligence and Robotics | Computer Sciences | Graphics and Human Computer Interfaces | Other Computer Sciences

Copyright

© Jeniya Sultana

Download

Available for download on Thursday, July 01, 2027

Open Access

Share

COinS