Course Description

This is an advanced course in natural language understanding. It focuses on methods for learning linguistic representations, at all levels of analysis: words, syntax, semantics, and discourse. The methods will mainly be probabilistic models, and it will emphasize the use of modern deep learning techniques in their design. We will cover applications of these models in a range of natural language understanding tasks such as language modeling, part-of-speech tagging, parsing, question answering, semantic parsing, sentiment analysis, and discourse coherence.

Prerequisites

This is not an introductory course. You must have taken and passed previous courses in these two areas.

  1. Natural language processing: ANLP or FNLP.
  2. Machine learning: IAML, MLPR, or MLP. If your only machine learning course is MLP and you are taking it concurrently with NLU, you should be very comfortable with it after semester 1.

If you’ve taken these prerequisites and done well in them then you will be well-prepared for NLU, and if you’re excited about these areas then I encourage you to take it.

If you have not taken previous courses in both natural language processing and machine learning, or if you took them but did poorly or didn’t enjoy them, then you should not take NLU.

Time and Place

  • Tuesdays 16:10–17:00, David Hume Tower, lecture theatre A
  • Fridays 16:10–17:00, Medical School, Teviot lecture theatre, doorway 5

Lectures will be recorded and made available through the university’s lecture capture service.

Teaching Team

Assessment

Assessment will consist of:

  1. A coursework, due in week 8, worth 30%. I encourage you to work in pairs—you’ll learn more that way.
  2. A final exam in the April/ May diet, worth 70%.

The course will follow the school-wide late coursework policy. To request an extension on your coursework, contact the ITO. If you contact me, I will simply redirect you to them, because I am not empowered to grant extensions myself.

The course will follow the academic conduct policy. I take this policy seriously: in 2017 I referred 22 students for academic misconduct, including collusion and plagiarism. I don’t enjoy doing this, but I will do it without hesitation if you violate the policy, because academic misconduct disadvantages other students. So please familiarize yourself with the policy.

Coursework

Due 16 March at 4pm (right before lecture).

Instructions

Code and data

Feedback and help

Through the course, you will receive feedback in several ways:

  • Short, non-assessed quizzes, to test your understanding.
  • Feedforward sessions for the coursework, where you will can ask questions.
  • Coursework assessment within two weeks of the due date.
  • A sample solution for the coursework, released after the deadline.

You can also seek additional help. We offer:

  • A weekly in-person office hour Fridays 2-3, starting week 3. Location TBD.
  • A weekly online office hour, Wednesdays 3-4, starting week 3. Details TBD.
  • A piazza forum. The teaching team will monitor the forum for unanswered questions once a day, Monday through Friday, usually in the afternoon. You are encouraged to answer other students’ questions, since if you can answer someone’s question, you know the material. We will correct any misunderstandings during our daily review.

If you have a question, please use one of these methods! Don’t email me—I recently crossed the email event horizon, so your email is likely to languish in my inbox for a long time before I see it. You will get a much faster response using the methods above.

Course catalogue


Informatics Forum, 10 Crichton Street, Edinburgh, EH8 9AB, Scotland, UK
Tel: +44 131 651 5661, Fax: +44 131 651 1426, E-mail: school-office@inf.ed.ac.uk
Please contact our webadmin with any comments or corrections. Logging and Cookies
Unless explicitly stated otherwise, all material is copyright © The University of Edinburgh.