2023–24 Projects:
At the home page for the Mercury Project at MIT's Laboratory for Computer Science, you will find a transcript of a dialogue between a person and the Mercury air travel planning system. Here is a portion of that dialogue.
USER: I WOULD LIKE TO FLY FROM BOSTON TO LONDON Mercury: Flights from Boston to London. What date will you be travelling? USER: NEXT FRIDAY IN THE EVENING Mercury: Okay. flights from Boston to London departing between 5:00 pm and midnight on June 25. Can you provide a departure or arrival time? The earliest nonstop flight from Boston to L ondon departing between 5:00 pm and midnight on June 25 leaves at 5:30 p.m. and arrives at 4:55 a.m. on British Airways. You arrive one day later. USER: IS THERE A UNITED FLIGHT Mercury: Okay. United flights from Boston to London departing between 5:00 pm and mid night on June 25. One moment please... I have a nonstop United flight leaving at 6:40 p.m. and arriving at 6:25 a.m.? USER: BOOK THAT FLIGHT Mercury: United flight 998 has been added to your itinerary. You depart Boston at 6:40 pm and arri ve in London Heathrow at 6:25 am. You arrive one day later. Where would you like to go next?
The Mercury system does not assume perfect speech recognition, which is good, since even human listeners don't do that. Instead, Mercury uses a sequence of prompts and verifications to make sure the system does what the user intends.
The Mercury system combines modules for speech recognition, natural language generation, text-to-speech, dialogue planning, database access, and so on. Much of the architecture of this system has been placed into an open source package called Galaxy Communicator. MIT has not released all portions of the Mercury project, but there are several other projects around the country that use the Galaxy architecture. Thus, there is a large collection of open source tools, based on a maturing architectural standard, for constructing special-purpose dialogue systems. In addition, some proprietary speech systems (Microsoft's SAPI, for example) include free tools for software developers.
Your goal for this project will be to assemble a dialogue system for answering simple questions about Carleton courses and registration. Your team will need to tend to several tasks, including:
In cases where it is appropriate and legal to do so, I have posted PDF versions of these papers in /Accounts/courses/comps/jondich under the last name of the first author (e.g. mctear.pdf).
Michael F. McTear, "Spoken Dialogue Technology: Enabling the Conversational User Interface," in ACM Computing Surveys, Vol. 34, No. 1, March 2002, pp.90-169.
The Mercury Project, Laboratory for Computer Science, MIT.
The Galaxy Communicator project.
The home page of SIGdial, the Special Interest Group on Discourse and Dialogue of the Association for Computational Linguistics.
Barbara Grosz, Discourse and Dialogue, Chapter 6 of Survey of the State of the Art in Human Language Technology, edited by Ronald A. Cole, et. al., 1996.