Text Analytics

Text Analytics: Deep Learning for Chatbots

Seminar kick-off on 18.04 between 8:00-9:40 at S202/C110

Course Description

Have you ever wondered the technology behind your Siri or ever wished your conversation with your Alexa were more engaging? In this seminar, we will take a closer look into how these “modern” conversational agents work and which deep learning techniques are used to build them. We will then discuss other fundamental challenges in the building process: available datasets (where do we train our models on?) and the evaluation techniques (how do we decide if a dialogue is 'good'?). Finally, we will investigate the answers to the question: 'what makes a good conversation?'. In light of these answers, we will discuss the shortcomings of the current models and the recently proposed models to make a 'better' conversation. By the end of this seminar, you should have a solid understanding of the recent advances in deep learning, how they are applied to build conversational agents and what is still missing.

Teaching Staff

  • Prof. Dr. Iryna Gurevych
  • Dr. Gözde Gül Şahin

We do not have fixed office hours. Please register via email if you need an appointment.

Literature

Will be announced during the seminar.

Timetable

The first sessions will consist of introductory lectures to cover the basics of machine learning methods used to build conversational agents. The program for the remainder of the seminar will be determined according to the number of participants and will answer the following questions:

  • What are the types of modern dialogue systems and what are they composed of?
  • How are they built? (Recent deep learning methods and available resources (tools, datasets) to build such systems)
  • What makes a good conversation ? (Evaluation of dialogue systems)
  • What will the chatbots look like in the future ? (Current limitations and the proposed methods to address them)