Spring 2019

We want you to get your hands dirty with most of the core topics covered in the course. To that end, we prepared two projects.

Groups: check Canvas or our blog posts.

Project 1

Deadline: April 26th, 2019

In this project you will learn about lexical alignment, the task of learning correspondences between words in different languages. You will apply latent variable modelling techniques, in particular, learning with directed graphical models (aka locally normalised models). In this project, you will parameterise the model using categorical distributions. You will experiment with maximum likelihood estimation.

Note: in IBM2 experiment use relative jumps (jump function).

Resources:

Submission:

TBA

Assessment: guidelines / grades on Canvas.

Project 2

Deadline: May 20th, 2019

Extension Deadline: May 28th, 2019 new!

In this project you will learn and implement a deep generative language model.

Resources:

Assessment: described in the project description / grades on Canvas.