Prè-requis
Basic linear algebra, calculus, probability theory, basic machine/deep learning
Objectif du cours
Speech and natural language processing is a subfield of artificial intelligence used in an increasing number of applications. This course will provide an overview and details of techniques and tasks used in the automatic processing of text and speech, covering certain history aspects of the field, the representation of textual and speech data, language modelling, machine translation, sentiment analysis and other labelling tasks, chatbots and speech synthesis and recognition. The aim is to provide the key principles, algorithms and mathematical principles behind the state of the art, and confronting them with the reality of processing real data.
En savoir plus : https://github.com/rbawden/MVA_2026_SNLP
Organisation des séances
The courses consist in 9 three-hours slots.
Each three-hour slot will have a lecture lasting approximately two hours, followed by a quiz and Q&As.
Mode de validation
Evaluation consists of 2 parts:
- Quizzes (40% of the total grade): For all classes except from the first and last classes, you will be given a link to an online questionnaire (google form) and will have 30 minutes to complete the questionnaire at the end of the lecture. Any forms submitted after the form is closed will be automatically rejected and graded as zero. The quizzes will contain comprehension questions and the best 6 grades out of the 7 quizzes will be used for the average.
- Final exam Project (60% of the total grade): You will work in small groups of 2-4 around a recent paper in speech or language processing which has existing code. But you will be evaluated individually on the basis of your engagement in the group work. Your task will be 1) to replicate the main result of the paper, 2) run an experiment testing a new question not tested in the paper (propose another approach/improvement to the task, or make the method work on another task or another dataset). You will present your plan in a one-page document for an initial deadine (TBD) (specify what you want to do, the planning, and how you will collaborate within the group), and your results in a 4-page document (deadline TBD) and 10-minute oral presentation (+ 5 minutes of questions) – time slots TBA.
Références
The recommended, but not obligatory textbook for the course is D. Jurafsky & J. Martin – Speech and Language Processing, 3rd (online) edition for already available chapters [J&M3], 2nd edition otherwise [J&M2]. Readings for each of the sessions will be provided by the instructors.
Thèmes abordés
Topics:
- speech features & signal processing
- hidden markov & finite state modelling
- word embeddings
- deep learning for NLP (RNNs, transformers)
- neural language modelling, including large language models (LLMs)
- machine translation
- sentiment analysis
- sequence labelling tasks
- chatbots
- evaluation: comparing human and machine performance
- speech synthesis and speech recognition
- code generation and agents
En savoir plus
Chloé Clavel
(INRIA)
Benoit Sagot
(INRIA)
Emmanuel Dupoux
(INRIA)
Rachel Bawden
(INRIA)
