CS 599 Dialogue Modelling  Spring 2004


Assignment 2: Building/Extending simple dialogue systems
Due: Monday February 16th

Topic: Movies of 2003

Since the oscar nominees were just announced, and we live in the film capital, it seems appropriate to build dialogue systems that can talk about the movies. Your assignment is to  build or extend three simple dialogue systems, using simple techniques.


Part I: Movie-chat
 The first part is to build a chat-bot that can have casual conversations about movies. You should modify Eliza or some other chatbot so that it can recognize comments and questions about movies and provide similar comments and questions. Some ideas:

What is your favorite movie?
I liked Lord of the rings.
So did I, but I hated Kangaroo Jack.
Did you see daredevil?
Yes, but the fights were better in the Matrix reloaded.
My favorite actor is Jack Nicholson.
Did you like him in Anger Management?
No, but he was great in "Something's gotta give".
I didn't see that. My favorite movie this year was "the last samurai"

Your modification should involve at least three different rules that can support a number of different inputs, and produce appropriate replies in some cases.

Turn in:
1) a listing of your changes to the code
2) a transcript of interaction with the system
3) (if possible) a  runnable demo


Some chatbot links:

• Lee Spector's simple eliza lisp code
• Chatterbot challenge contestants (some have source code)
• Simon Laven's chatterbot page
  

Part II: Oscar Prediction Taker - finite state system

Design a finite state dialogue system that can give information about nominated movies and accept votes
for each category. Include at least the categories: Best Picture and Best Director.
People should be able to perform the following tasks:

a) get  total number of nominations and director  for each film
b) for each director nominated, get the film they directed
c) get all the nominations for each category
d) vote for a specific nominee for the poll
e) get a tally of the leading nominee for each category


Turn in:
1) a snapshot or diagram of the network
2) a sample transcript illustrating each of the five types of tasks
3)   a runnable demo


Use the RAD



Part III: Oscar Prediction Taker - voice xml
 
Produce a system with the same coverage as in Part II, only using voice xml.

Turn in:
1) vxml code
2) a sample transcript



Voice XML links:

• Elvira  free Voice XML interpreter
• ken Rehor's World of VoiceXML
• Dave Ragget's tutorial
  

Part IV: Analysis

Write a page or two outlining your approach and experience with each system. What are the strengths and challenges of each approach?
What kind of applications (if any) would you prefer to use each platform for?