Seminar Chatbots and Conversational Agents

Nowadays, chatbots (or conversational agents) are available on different platforms. It can be in a professional context (Skype, Slack) or a private one (Facebook Messenger, Discord, Telegram). Since Facebook’s F8 2016 conference, these bots have become more democratic and have flooded the different platforms. They tend to replace mobile applications for their ease of use and various functionalities. In addition, there is no need to install them because they are available on apps that everyone already has.
Different uses stand out. For example, it is a way for brands to automatically answer questions from users (support, after-sales service). These bots can work as assistants for different tasks (make appointments, various automations, reminders). Online stores use it to allow people to place orders as if talking to a person. The news sites use them to distribute a summary of the articles of the day.

Transactional chatbots are all over the place and are the first step of chatbots. We now search for more human-like conversations, with emotions, understanding and eventually empathy. We want to coach the user, to understand him, to help him, depending on the context (eg. customer service, mental illness treatment, coach, companion). Thanks to the outstanding evolutions in artificial intelligence and natural language processing in the last decade, it is now possible to converse in a more or less natural way with bots.

This seminar focuses on investigating approaches and technologies used in today’s chatbots, as well as comparing and applying underlying psychological theories. There are different types of bot for different uses. It is then worthwhile to understand the advantages and disadvantages of each of these techniques. All bots are not powered by machine learning, some are rule-based. Natural language understanding as well as natural language generation are two fundamental aspects of chatbots. Monitoring a context in a conversation is also a very important element. Analyzing and designing a human-machine interface is not an easy task. Bots can have moods and a personality. All these aspects must be considered while designing a chatbot.

A particular emphasis this year will be given to mood, personality and language generation.
The seminar will have a practical component as students will investigate existing chatbots as well as develop new concepts in the aforementioned domains.

Details

Code 33798
63798
Type Seminar
ECTS 5
Site Fribourg
Track(s) T3 – Advanced Information Processing
T6 – Data Science
Semester A2019

Teaching

Learning Outcomes
  • Identify and describe existing approaches and mechanisms for designing chatbots
  • Identify the main components of a chatbot architecture
  • Discuss and compare the different techniques available for chatbots with their strengths and weaknesses
  • Compare and identify the best technological solution to design a specific type of chatbot
  • Understand the techniques used to follow the context in a conversation
  • Know the landscape of existing bots in the nutrition domain
  • Evaluate the capabilities and skills of a bot
Lecturer(s) Elena Mugellini
Omar Abou Khaled
Philippe Cudré-Mauroux
Language english
Course Page

The course page in ILIAS can be found at https://ilias.unibe.ch/goto_ilias3_unibe_crs_1531295.html.

Schedules and Rooms

Period On Appointment
Location UniFR, PER21

Evaluation

Evaluation type continuous evaluation

Additional information

Comment

First Lecture
Please contact Elena Mugellini (Elena.Mugellini@hefr.ch) to select the date of the first meeting.