PythonNLP

What to expect:

This is a more advanced NLP course for linguists or CS majors new to NLP. It will cover the basic types of ML algorithms that are widely used in the current NLP, striving to accommodate even the attendees without mathematical background. Additionally, it will discuss advanced NLP topics in which linguistic expertise of ESSLLI attendees is needed the most: the creation of datasets, evaluation, and general methodology.

We will also discuss the latest developments in word-level, compositional and contextualized meaning representations.

Pre-requisites:

Basic programming concepts and some familiarity with Python (variables, loops, functions), such as can be obtained from this free introductory MITx course.
Familiarity with the field of NLP and basic machine learning concepts (covered in the week 1 of the course)
Prior knowledge of calculus and linear algebra is encouraged, but not required. The linguists in the audience will be provided with the best possible overview of what's going on conceptually, and the countless possibilities of useful research they can do with the models already published by others.

Lectures will include technical sessions (Machine Learning), hands-on Python tutorials with PyTorch, and theoretical overview sessions.

Lecture 1 (Monday)

ML: Logistic regression, classification, regularization. [slides]
Tutorial: Sentiment analysis with logistic regression in sklearn. [tutorial]

Lecture 2 (Tuesday)

ML: Introduction to deep learning, Feed-forward neural networks. [slides]
Tutorial: Sentiment analysis with feed-forward neural networks in PyTorch. [tutorial]

Lecture 3 (Wednesday)

ML: Word embeddings [slides]
Tutorial: Training, exploring and using word embeddings [tutorial]

Lecture 4 (Thursday)

Lecture: Evaluation of word embeddings [slides]
Tutorial: Profiling word embeddings with LDT [tutorial] [corpus resources] [language resources]

Lecture 5 (Friday)

Lecture: NLP datasets: methodology and current challenges (continued) [slides].
Closing remarks [slides]

Any feedback for the course is much appreciated! [Feedback form]

LDT library notice: you can sign up for the mailing list to get notified of updates. The fix for the Windows installation problem is forthcoming.