The course can be taken as (i) an individual reading course; (ii) a seminar course with other students; (iii) partially overlapping with a corresponding masters course; or (iv) a combination of the previous. Participation in summer schools (for which you have not already obtained credit) may also be counted here.

This means that each course participant will have an individual plan of activities. Please make sure that you get in touch with the course organiser(s) as soon as you decide to take this course to agree on a plan. Subsequent integration of activities may not always be possible.

Individual course

The course is organised as individual discussions based on your readings and practical implementations of solutions from papers. We will jointly choose a list of topics based on your interests and needs and for each topic we will agree on a reading list or a list of other resources (e.g. datasets and code) that will be considered. For each topic you should prepare written work (a summary of readings, a practical programming implementation or experimental results) that will then be discussed in a meeting. To get an overview of subject (and to pass the course) you should look at least 6 topics. You should also complete a course project which should be of a quality of a workshop paper that could be later submitted to a workshop or a conference.

Seminar course

If more students are taking the course, it will run as a seminar series/reading group. We will jointly choose a list of topics based on students’ interests. In order to pass this course, each student should participate in at least 12 seminars for which they should read a paper (or look at code and other resources) and participate in discussions, lead a discussion in least 4 seminars, and complete a course project which should be of a quality of a workshop paper that could be later submitted to a workshop or a conference.

With the masters course(s)

If this course has a corresponding masters course within the MLT programme and you have not taken it yet before, you can also follow its lectures, tutorials and labs. In order to pass this course, you should attend the lectures, tutorials and labs of a corresponding masters course, pass all its labs and complete a course project. Doing so you should exemplify a higher quality and sophistication of work, appropriate for the PhD level. For example, the project should be of a quality of a workshop paper that could be later submitted to a workshop or a conference.

Mix and match

A combination of the above options is also possible. Here are some examples:

Note that if you are taking more than one course a particular piece of work can only be considered for one course.

How much work is required?

An individual plan will be agreed with each student at the start making sure that equal amount of work is required to pass the course whichever option is followed. Generally, as this is a 7.5 HEC course it means that you would spend working on it for 5 weeks full-time or 10 weeks half-time (60 HEC is 40 weeks).

Submitting your work

Because you have an individual course plan it is important that you keep a good record of what you have done. No record and materials to show your work - no credit! It is also recommended that you do this as you go along and check with the teachers what counts for what course and not wait until the end when you might already forget or loose important code and notes or it may happen that the work is not related to the course content.

Therefore, for each course that you are taking, create a private Github repository and add Simon (sdobnik) as a collaborator as soon as you start with the course.

Then create the following files and folders as needed:

  • README.md: Contains your individual course plan and a description of options that you taking or have taken. Please update this file regularly as you complete new parts.
  • reports: Create a sub-folder for each of the topics and include a written report with references, documented code and other relevant materials.
  • For seminars/reading groups, list the seminars that you attended with references to papers that were discussed and whether you have lead the discussion in the main README.md.
  • project: A home for a well documented code (e.g. as a Jupyter notebook) and the final course paper with references. Use a separate cloud repository for large files such as datasets and models and include the link in the paper. The paper should also include a description of how the code should be read. Use the latest ACL/NAACL/EMNLP template.
  • library: a folder for any background supplementary resources and information;
  • ...

Grading

The course is graded with G (pass) and UG (fail).