This course provides participants with the theoretical foundations and practical methodologies required to generate synthetic textual data using Generative Artificial Intelligence. Participants will learn how synthetic datasets can support the development, testing, and validation of Machine Learning and Natural Language Processing models, while gaining awareness of the quality, ethical, and governance aspects associated with synthetic data generation.
By the end of the course, participants will be able to: Understand the theoretical foundations and motivations behind the use of synthetic data; Analyze the main approaches and technologies used for text generation; Design synthetic datasets for machine learning training and testing activities; Evaluate data quality, ethical implications, and risks related to synthetic data usage.
Course Content and Programme
Module 1 – Foundations of Synthetic Text Data Generation (1,5 hours)
Provide participants with a comprehensive overview of modern Generative AI technologies for textual content generation and their application to synthetic dataset creation.
- Introduction to textual data and its main typologies.
- State-of-the-art technologies for text generation.
- Transformer architecture and its evolution.
- Large Language Models (LLMs), including GPT, LLaMA, Qwen, Gemma, and related architectures.
- Prompt engineering techniques.
- Text controllability and guided generation.
- Conditional generation based on labels, styles, and textual structures.
- Introduction to LangChain and LLM-based pipelines.
Module 2 – Quality Assessment and Governance of Synthetic Data (0,5 hours)
Provide participants with criteria and methodologies for evaluating generated data and understanding associated risks.
- Quality dimensions of synthetic textual datasets.
- Data consistency, diversity, and representativeness.
- Ethical considerations in synthetic data generation.
- Bias, privacy, transparency, and governance challenges.
- Risk assessment and mitigation strategies.
Module 3 – Hands-on Laboratory: Building Synthetic Text Datasets (1 hour)
Apply the concepts introduced during the course to the design and generation of synthetic textual datasets.
- Practical workflow for synthetic dataset creation.
- Configuration and use of generative AI tools.
- Dataset generation strategies for training and testing purposes.
- Quality verification and optimization techniques.
Course Duration
3 hours
Target Audience
The course is aimed at:
- Data Scientists;
- Software Engineers;
- Researchers and Research Technologists;
- AI and Machine Learning Developers;
- Master’s students in Computer Science and Artificial Intelligence;
- NLP Engineers and professionals working with language technologies.
Requirements
Participants should have:
- Basic knowledge of Python programming;
- Fundamental understanding of Machine Learning concepts;
- Basic familiarity with Natural Language Processing techniques.
Calendar
October 28th, from 10:00 to 13:30