Course synopsis and format
This course aims at providing basic knowledge about NLP from three perspectives:
- linguistic background for engineers;
- models and algorithms (technology);
- typical tasks/applications.
Linguistic background for engineers
- Linguistic levels (lexical, syntactic, semantic, pragmatic)
- Impact of power laws
- Word morphology
- Syntactic parsing
- Lexical semantics and semantic relations
Models and algorithms
- Evaluation methodology and metrics
- Estimation and smoothing techniques
- CYK algorithm
- Naive Bayes classification and Multinomial Logistic regression
- Vector-Space model
- word2vec and GloVe
- (intro to) transformers
Tasks and Applications
- Machine Translation
- Word-Sense Disambiguation (WSD)
- Language Identification
- Spelling-error Correction
- Named Entity Recognition (NER)
- Part-of-Speech tagging
- Text Classification
- Information Retrieval
- Sentiment Analysis
Detailed content with weekly schedule of review lectures, hands-on, exercises and practical sessions is provided on the companion website coling.epfl.ch
Course FormatWeeks 2 to 11 of the course are "flipped-classroom": the lectures will be provided, prior to official time-slots, in the form of videos (links to appear on coling.epfl.ch and all videos available from there), and we will spend 2 hours in class to comment the lectures, discuss them with you, provide hands-on and discuss yours results,etc.
Then 2 hours of exercises/practice (with TA) and free-questions (1-on-1 with teacher, pretty much like "office hours"), will follow.
Nothing done in class is recorded (although broadcasted via Zoom) as it'll be too much material and, mostly, will miss the main goal of these sessions: live contact, interaction, adaptation, etc.
So, if you don't come to the "lectures"/exercises, you'll miss the contact, but you won't miss any major content (pretty much as if you were studying with books only, without coming to the lectures). But, of course, attending the contact hours is highly recommended.
For asking questions asynchronously, you can use the discussion forum on this Moodle site.
Notice also that the solutions of the hands-on, the exercises and the practical session will be provided (weekly).
For the last 3 weeks of the semester: it'll likely be given in person but recorded on the spot and made available online after the lecture.
Regarding the assessments: 16% of the grade will be done during the semester in the form of four on-line quizzes (mandatory presence in class; dates provided on coling.epfl.ch) and a 84% final exam (written), during the exam session.