NLU-11: Week 7 | Open Course Materials

This week, we will look at evaluating generation and machine translation. We will also look at multilingual machine translation and how this is different when using large language models.

Lab 3 takes place this week. Please have a look at the lab exercises before you attend the lab session, especially at Section 1:

Lab 3: Tensor Computation in PyTorch

This lab teaches tensor computation, batchification and masking in PyTorch.

Coursework 2 has been issued and is due on Friday 21 March at 12:00. Please bear in mind that some aspects of the coursework will be time consuming to run. Start working on the coursework early! The question sheet contains more guidance on this, please read it now.

Here are the solutions for last weeks tutorial:

Solutions for Tutorial 2

Slides	Lecture	Course Content
Evaluating Translation and Generation [handout]	19	Required reading: Bleu: a method for automatic evaluation of machine translation, Papenini et al. (2002) (Section 1-3) Additional reading: COMET: A neural framework for MT evaluation, Rei et al. (2020) (sections 1-2) Bertscore: Evaluating text generation with bert. Zhang, Tianyi, et al. (2019) (sections 1-3) ROUGE: A Package for Automatic Evaluation of Summaries, Chin-Yew Lin (2004) (sections 1-4)
Machine Translation and Multilingual data	20	Required reading: Multilingual Denoising Pre-training for Neural Machine Translation Liu et al. (2020) For interest, not require reading: Survey of Low-Resource Machine Translation, Haddow et al. (2021) No Language Left Behind (Costa-jussà et al. 2023)
Translation and LLMs	21	Optional Reading: How Good Are GPT Models at Machine Translation? A Comprehensive Evaluation Hendy et al. 2023 A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models Xu et al (2023)

License

Search

Navigation