Week 10: Co-reference
Reminders and announcements
- Assignment 2 is due on Wed at noon.
- There are two help hours remaining before then in AT 5.07: Mon at 10am and Tue at noon.
- No extensions are permitted on this assignment and late submissions will receive a mark of 0, so plan on finishing early just in case! Also remember: it is much better to submit an incomplete assignment than to submit late or not at all.
- Submission instructions are the same as for Assignment 1: see the instructions at the bottom of the Week 5 page.
- Tutorial 4 meets this week. Before you look at the questions, please read below.
- The two questions are both from past exam papers, so if you haven't already, we recommend you read through the exam revision guide before attempting the questions.
- We recommend you treat this as a practice exam: set a timer for 45 minutes from when you first look at the questions, and write your answers on paper.
- Each of the two tutorial questions was originally worth a total of 10 marks (out of 50) in a past exam (these were questions from "Part 2" of those exams, as described in the Revision Guide). The questions span material from the whole semester, and will test your knowledge of multiple topics.
Overview of the week
Welcome to Week 10, the final week of the course! It simultaneously feels like a very long time and just yesterday that we started this course together. I hope you will all take a moment to congratulate yourselves on making it this far, and to reflect on what you have learned so far.
This week the examinable lectures will be focusing on co-reference, covering some of the linguistic information that is relevant for the task, and also some of the ethical issues that arise (namely, gender bias). We'll discuss one way of trying to measure some of those biases, which is potentially relevant in other areas of NLP as well. We will also have a guest lecture from Emily Allaway, a new member of academic staff, who will discuss implied meaning in text and related applications such as stance detection and misinformation/misreprentation detection.
Finally, we know many people are feeling pressured with assignments this week, but we would strongly encourage you to set aside a quiet hour to prepare for your tutorial (15 minutes to read the exam guidance, and no more than 45 minutes to write your answers to the questions), and then attend the group meeting! This is one of your best opportunities to discuss and get feedback on your answers to exam questions that span the whole course.
Lectures and reading
Lecture # | Who? | Slides | Reading |
---|---|---|---|
1 | SG | Co-reference | JM3 23.0-23.1 (*), 23.2 |
2 | Guest lecture (Emily Allaway) | Implied meaning in text | none |
3 | SG | Gender bias and challenge datasets | JM3 23.9-10 (*) |