Eli Pincus, pincus [AT] ict.usc.edu
This course will provide an overview of natural language dialogue systems. Natural language dialogue systems allow human users to interact with computers using a natural language such as English (spoken or written). This human-machine interaction takes place in the form of conversation. Natural language dialogue systems have many applications, for example, providing information, tutorial dialogue (and generally dialogue for training, e.g., leadership skills, interviewing skills, etc.), healthcare applications (e.g., virtual human interviewers that interact with people who suffer from depression or post-traumatic stress), storytelling, controlling smart homes, companions for the elderly, etc. Recently, conversational assistants (a special type of natural language dialogue systems), e.g., Apple Siri, Amazon Alexa, Google Home, Microsoft Cortana, have become very popular. Throughout the course, students will learn about the capabilities and limitations of such systems.
The topics covered will be similar to CSCI 644 Fall 2018 including (but not limited to):
Half of the course lecture period will be a lecture by the instructor (or guest instructors) and half will be group discussion of research papers related to a specific topic (usually the topic presented in the first half of the lecture period). These research papers will be selected before the lecture and students will have to read and review them and send in discussion questions. Group discussions will be led/co-led by one or more students. Discussion leaders will start the discussion with a short presentation on the topic.
Students will also have to complete 2 assignments, and carry out a main project. For the assignments, students are expected to work separately. For the main project, students can select their own topic and work separately or form groups. The topic of the main project should be relevant to the course and approved by the instructor. Because the field of natural language dialogue systems advances rapidly and the state-of-the-art is continuously changing, there is no required textbook. The material will be covered through lectures and assigned readings.
The schedule below is TENTATIVE.
| Date | Speaker | Topic |
| January 17 (F) | Georgila | Overview of natural language dialogue systems |
| January 24 (F) | Georgila | Basic principles of dialogue processing
Knowledge-based dialogue management |
| January 31 (F) | Georgila | Explanation of Assignment 1
Speech recognition for dialogue |
| February 3 (M) | Preferences for papers (to lead discussion) due at 1pm | |
| February 7 (F) | Georgila | Conversational speech synthesis |
| February 14 (F) | Assignment 1 due at 10am | |
| February 14 (F) | Georgila | Natural language understanding for dialogue (and dialogue state tracking)
Natural language generation for dialogue Student presentations/discussions #1 Topics/questions for discussion should be submitted on Blackboard, due February 13, 1pm |
| February 21 (F) | Georgila | Reinforcement learning approaches to dialogue and simulated users (part 1)
Project discussions |
| February 28 (F) | Georgila | Explanation of Assignment 2
Student presentations/discussions #2 Topics/questions for discussion should be submitted on Blackboard, due February 27, 1pm |
| March 3 (T) | Project proposals due at 10am | |
| March 6 (F) | Georgila | Reinforcement learning approaches to dialogue and simulated users (part 2)
Student presentations/discussions #3 Topics/questions for discussion should be submitted on Blackboard, due March 5, 1pm |
| March 13 (F) | Georgila | Dialogue system data collection and evaluation
Student presentations/discussions #4 Topics/questions for discussion should be submitted on Blackboard, due March 12, 1pm |
| March 20 (F) | Spring break (no class) | |
| March 27 (F) | Georgila | Incremental dialogue processing (and turn-taking)
Student presentations/discussions #5 Topics/questions for discussion should be submitted on Blackboard, due March 26, 1pm |
| April 3 (F) | Guest lecture Prof. David Traum |
Story, identity, and relationships in human-machine dialogue
Student presentations/discussions #6 Topics/questions for discussion should be submitted on Blackboard, due April 2, 1pm |
| April 10 (F) | Guest lecture Dr. Mark Core |
Dialogue systems for education
Student presentations/discussions #7 Topics/questions for discussion should be submitted on Blackboard, due April 9, 1pm |
| April 13 (M) | Assignment 2 due at 10am | |
| April 17 (F) | Georgila | End-to-end dialogue systems and chatbots
Student presentations/discussions #8 Topics/questions for discussion should be submitted on Blackboard, due April 16, 1pm |
| April 24 (F) | Project presentations (Part 1) | |
| May 1 (F) | Project presentations (Part 2) | |
| May 6 (W) | Project report due at 4pm |