Speech recognition and kitchen computing: Speech recognition is becoming increasingly ubiquitous, even in the setting of the household kitchen. This project involved developing a speech interface for an intelligent, internet-capable refrigerator. This included maintaining a grocery shopping list, a recipe transcriber, and an interactive recipe engine. This helps create a safer and more convenient environment for meals preparation and other kitchen tasks.

Dental corner after-hours automated appointment service: This project involved developing an automated system for making, checking, changing and cancelling appointments by phone for a dentist's office. It used the PHP scripting language to create dynamic VoiceXML pages. Hinges were developed to permit information flow between the application and an external database containing information relevant for scheduling apppointments.

An interactive driving directions-finder: This project involved development of a hands-free, speech-based directions finder to assist vehicle drivers. Based on the OGI Toolkit RAD canvas, it allows the driver to ask for a destination and specify the current location. The system then uses the HTTP Tcl package to access Web at maps.yahoo.com and retrieves the relevant directions to the destination. It parses out the textual form of the directions, normalizes the content (e.g. replacing many abbreviations with full words), and then sends the result through a TTS engine.

Foreign language recognition system: This project involved development of a Java-based distributed foreign language recognition system. It is implemented as a distributed processing telephone-like IP Multicast system in which a user connects to the network and begans talking; a number of operators compete with each other by a series of bids with the task of recognizing the user’s language.  The winning operator then forwards the client’s call to someone capable of talking with the client. The language identification core is not phoneme-based, but rather speech-vector driven, based solely on the acoustics qualities of the waveform. Features are extracted, normalized, and categorized with neural network-like approach. The general framework was successfully applied in distinguishing English from Vietnamese.

A voice interface to the scriptures: This program allows for speech-based retrieval of scriptures. It combines the OGI toolkit plus Tcl-based extensions as well as XSLT style sheets and a modified XML-marked-up version of the Book of Mormon. The program prompts the user to give a scripture reference, and then uses the words recognized to generate an XSLT transform. The XSLT is fired and returns the verse in a file which the toolkit's TTS system reads back to the user.

GedSpeak: speaking to your genealogy: This project enables users to access genealogical information in a hands-free fashion via speech. The work involved designing, implementing and testing a speech application designed to enable a user to query information contained in GEDCOM files. The widely used OGI speech toolkit with its RAD dialogie development canvas was used, along with custom Tcl code to perform queries in GEDCOM databases to respond to user queries.