A speech-enabled system for website bookmarking

Abstract

In recent years, much advancement has been made in both search and speech technology. The former seeks to organize and retrieve the ever-growing amount of online information efficiently, while the latter strives to increase mobility and accessibility in multimodal devices. Naturally, searching via spoken language will become ubiquitous in the near future. As a step towards realizing this goal, this thesis focuses on the simpler problem of bookmarking and retrieving websites via speech. With data collected from a user study, we gained insight on how to predict speech tags and query utterances based on a webpage's content. We then investigate and evaluate several heuristics for tagging and retrieving bookmarks with the objectives of maximizing recognition accuracy and retrieval rates. Finally, the progress culminates in a prototype Firefox extension that encapsulates an end-to-end system demonstrating speech integration into the bookmarking capabilities of the Firefox browser.