Lecturer: Raquel
Fernández (Institute for Logic, Language and Computation)
Teaching assistant: Dieuwke Hupkes
Timetable: see https://rooster.uva.nl (click on "Add Timetable > Course" and enter the name of the course)
Overview: This year the course will be dedicated to linguistic interaction (to models of language as it is used in actual conversation). This website will be updated throughout the course with the contents of each week. The information may be tentative and it is subject to change. Please check this website regularly.
Prerequisites: There are no formal prerequisites, but I'll assume some knowledge of semantics/pragmatics of natural language. Basic programming skills in Python are needed for homework assignments. If you don't have any programming kwnoledge, you are welcome to stay in the course provided you are willing to learn by working with a fellow student who has more experience.
Evaluation: The first part of the course will consist of lectures, practical assignments, and discussions of research papers; everybody is expected to play an active role in class. In the second part of the course you should work on a research project of your choice. You will be asked to write a short paper by the end of the course. This paper will account for most of the final grade (around 60%). The rest of the grade will correspond to the practical exercises and class discussions.| Time | Contents | Materials & Assignments |
|---|---|---|
| Week 1 |
We will spend the first week discussing turn taking, i.e., how dialogue participants coordinate the timing of their contributions in conversation.
06 Sep: Overview of the course and introduction to turn taking [SLIDES] 08 Sep: Discussion of two research papers: Helden & Edlund (2010) and Laskowski (2016) |
For a general, high-level overview of dialogue research, see the chapters on Dialogue and on Spoken Language Dialogue Systems in the Oxford Handbook of Computational Linguistics (2nd edition).
On the slides, you will find many references on the concrete topics we cover. |
| Week 2 | There are no lectures this week since Raquel is away at the Interspeech and SIGdial conferences (you may want to check their programmes - they are very related to the contents of the course).
13 Sep: Practical session run by Dieuwke, where you will get familiar with the environment and the dataset we use for the assignments, and can get started with Assignment 1, which is related to turn taking. 16 Sep, 11:00: deadline to submit Assignment 1. |
Here are the instructions for the practical session.. Assignment 1 is available on Blackboard and should also be submitted via Blackboard. |
| Week 3 | After having discussed timing coordination, we will move on to issues realted to content (aka meaning) and content coordination.
20 Sep: Brief discussion of Assignment 1 and introduction to dialogue acts [SLIDES] 23 Sep: Discussion of Milajevs & Purver (2014) on dialogue act classification and, if time permits, we will start to discuss the joint action model and its theory of grounding. [Time permitted going over a few slides] |
In Assignment 2, you will have the chance to explore the dialogue act annotation of the Switchboard corpus. The assignment will be available on Blackboard after the lecture on 20 Sep. |
| Week 4 |
27 Sep: More on grounding, specially in referential matching tasks [SLIDES], and discussion of Manuvinakurike et al. (2016) on incremental dialogue act segmentation.
30 Sep: Comments on Assignment 2, and discussion of Geertzen (2015). We then moved on to discuss style coordination [SLIDES] |
The submission deadline for Assignment 2 is Tuesday 27 Sept at 23:00. Please submit your Jupyter notebook file via Blackboard.
Assignment 3 on style coordination is now available on Blackboard. |
| Week 5 | 4 Oct: We will discuss two papers related to style coordination, Danescu-Niculescu-Mizil et al. (2012) and Reitter & Moore (2014) (since the latter is rather long, you may instead take a look at an shorter and more preliminary version: Reitter & Moore (2007)).
[SLIDES]
7 Oct: We will discuss first language acquisition in interaction. A good part of the lecture will be based on this recent paper: Hiller & Fernández (2016). [SLIDES] |
The submission deadline for Assignment 3 is Friday 7 Oct at 23:00.
Use the discussion forum for "Project ideas" on Blackboard to get feedback on your proposals and try to get a project partner! |
| Week 6 | 11 Oct: Following up on language acquisition, we discussed one of the outstanding papers at ACL this year, by Wang, Liang & Manning (2016) on language learning by an artificial agent. We also went over the next steps and the requirements for the final paper and the presentation.
[SLIDES]
14 Oct: Brief comments on assignment 3 (if applicable) and discussion of your project proposals. |
Submit a project proposal by Thursday 13 Oct, 14:00. See instructions on slides. Each of you will get a proposal by a fellow student by 16:00 on that day, and you will be asked to briefly explain it in class on Friday. Everybody is expected to provide feedback. |
| Week 7 | There are no lectures this week. Independent project work and personalised supervision. Book an appointment with Raquel no later than Monday 17 October. | Supervision meetings take place in Raquel's office, room F1.07 at the ILLC. Bring the current draft of your paper to the meeting.
|
| Week 8 | 28 Oct: Final project presentations, seminar room F1.15 (opposite Raquel's office at the ILLC): 10:30 arrival and setting up 10:35 Prajit & Jacob (20 + 5) 11:00 Loïs (15 + 5) 11:20 Eiseart (15 + 5) 11:40 break 12:00 Laura, Natalia & Valentin (25 + 5) 12:30 Peter & Victor (20 + 5) 12:50 end 31 Oct: Final project submission deadline (submit via Blackboard by midnight). |
Grading criteria for presentations and for final papers. Your project paper must be prepared using the Latex style files of EACL-2017. The template can be downloaded from here (direct download of zip file). If you are an Overleaf user, there is also an Overleaf template. When you prepare your paper, make sure you uncomment line 25 in the source tex file (\eaclfinalcopy) since your paper should of course not be annonymous. |