Natural language processing - KAIN8M13

Goal(s)

This course introduces engineering techniques used in the characterization and modeling of language information. Its objective is to introduce the processing techniques used in current language information analysis systems, based on textual and/or vocal data corpora. The objectives of this course are as follows:

• To know how to extract representations of acoustic and textual data in order to analyze or model these data
• To know how to implement a processing chain to model language information
• To know how to use large language models (LLMs) available on open-source platforms (Huggingface) for modeling tasks
• To develop a critical sense of analysis

Content(s)

1. Introduction: Communication, Language, and Technologies
2. Representation and Encoding of Speech Signals
3. Extraction and Encoding of Textual Data
4. Preprocessing and Language Data Models
5. Methods and Metrics for Evaluating Inferential Systems
6. Artificial Neuron for Machine Learning
7. Artificial Neural Networks
8. Advanced Neural Architectures (CNN, Transformers)
9. Self-Supervised Methods for Learning Large Language Models (Speech)
10. Self-Supervised Methods for Learning Large Language Models (Text)

Basic notions in signal processing, information theory and computer science.