Messages
In this post you will find the detailed program for the WNNLP 2021 workshop, which will be held on May 27th 12:00 - 14:00.
- 12:00 - 12:05 Welcome
- 12:05 - 12:10 Task overview: Natural Language Inference
- 12:10 - 12:15 Erik Andreas Johansson and Myrthe Lammerse "BERT for natural language inference on the MultiNLI and XNLI"
- 12:15 - 12:20 Sondre Wold "The Effect of Different Hyperparameters on BERT for Natural Language Inference"
- 12:20 - 12:25 Vilde A. Huseby and Bendik M. Aas "Cross-Lingual Natural Language Inference on XNLI using Multilingual BERT"
- 12:25 - 12:30 Daniel Clemeth Bjerkeland "Training BERT for Natural Language Inference...
WNNLP 2021 will be held on May 27th 12:00 - 14:15. Please see the course web page for a detailed program (coming very soon).
At the workshop, each accepted paper must be presented in a lightning talk of no more than 5 minutes. Below are some instructions on the presentation:
- For team submissions, please agree among yourselves on who will give the presentation. Since we have a very tight program there will be no time to switch between different presenters for one paper.
- Each track (TSA, NLI and WSI) will be presented by the task chair at the conference and will briefly introduce the task, data and evaluation metrics. The lightning talks for each paper should therefore focus exclusively on the experiments and results of the specific paper being presented.
Please remember to submit your camera ready papers by May 25th, 23:59 (CET). This is a strict deadline and there will be no extensions.
We are now ready to start the review process for WNNLP 2021. You should have by now received the notification by email. Please login to your Easychair account, which contains a list of papers assigned to you as a reviewer.
Your reviews are due Friday, May 14, 2021 (11:59pm anywhere on Earth).
Note that the 'exam' folder in the course GitHub now contains two sample reviews that you can use for inspiration.
Here are a few guidelines for the review process.
- Be timely: Even if you don’t plan to start your reviews as soon as they are assigned, please do log in to Easychair and see which papers you got. This will allow you to notify us of any issues in time for us to reassign the paper. Furthermore, please don’t count on finishing your reviews at the last minute. As we all know, things can come up, and...
You should have received an e-mail to join the WNNLP 2021 Program Committee as a reviewer. Following submission of your papers on May 6th (23:59, anywhere on earth) you will be assigned 2-3 papers to review by May 14th. Paper assignment will be completed by the end of May 10th. Please accept the invitation sent via e-mail to join the PC and get ready for reviewing.
We wish you the best of luck in wrapping up the home exam and submitting your paper for WNNLP 2021!
Here's the leaderboard for oblig. 3:
1. adriase, danish, haiyuec - 0.9116
2. erikajo, myrthel - 0.9091
3. nikolhm - 0.9013
Congratulations! You've all been given +1 point for the oblig.
All the obligatories have now been marked, and we have our leaderboard! Again, same caveats as last time apply - these random fluctuations towards the top are hard to explain and realistically all the models are probably pretty similar. Having said that, the leaderboard is:
- stiansgr, janilam, iversw - 0.825
- anettfre, nichow, anthip - 0.824
- adriase, danish, haiyuec - 0.817
Good job!
The third obligatory assignment is now online, on the Git repository. It is titled "Named entity recognition and pre-trained language models" and is due April 12.
Transformer architecture has already been covered in the lecture last week. Contextualized models will be covered in the lecture this Thursday.
We will provide extensive practice with pre-trained language models (including fine-tuning) at the group sessions.
Please check Mattermost or the readme in the git repo to find the link to the recurring Zoom meeting we will be using for the remaining lectures this semester.
The second obligatory assignment is now online, on the Git repository. It is titled "Word embeddings and Recurrent neural networks" and is due March 13.
Feel free to start working on the word embeddings part. We will begin covering recurrent neural networks (RNN) in the lecture next week.
Here's the leaderboard for the Obligatory 1 (macro-F1 scores):
1. fabior, pmhalvor, eivindgg - 0.47
1. markuhei, sondrew - 0.47
2. yauhenk - 0.45
Good job! Each of you gets an extra mark, if you aren't already at full marks.
The first obligatory is now online, on the Git repository. Feel free to start working on it, and bear in mind that things will become clearer with more lab sessions :)
Hi everyone,
The first IN5550 group session tomorrow will have to be held digitally (not surprisingly). Please find the Zoom meeting link in the IN5550 UiO Github repository or in the IN5550 Mattermost chat.
If you have any problems with getting access, make sure to drop us a message, and we will help you.
The group session will feature two main parts:
- Working with Saga
- Linear algebra with PyTorch
It is desirable for all the students to at least apply to the Saga account, so that we can sort out all technical issues during this first group session.
See you tomorrow in Zoom!
The first lecture in IN5550 this term will be held on Thursday, January 14, at 12:15. We will go through course logistics (including routines for assignments and the final project-based exam) and motivate the now dominant use of neural architectures in Natural Language Processing (and most other sub-fields of Artificial Intelligence). The first lecture will be virtual (via Zoom) and further details will be provided prior to the lecture on the course web page (this page).
To prepare for this class, we ask that everyone fill in a brief anonymous survey about their background and request access to the Saga national supercluster.
The current plan is for subsequent lectures to be held physically, but screencasts will be made available via the course web page following the lecture. Depending on the corona situation, t...