- Level: Basic - Choice - Specialization
- The essential course to understand the impact of data on interaction AI systems, and strategies for effective data collection and preparation.
- Prerequisites:Introduction to Interaction AI or any introduction course.
- Assessments: Micro Assessment weekly, Full Assessment last week.
Week 1 - Introduction to Data for Interaction AI
Learning OutcomeStudents will understand the different types of data necessary for building interaction AI systems and their respective preparation techniques.
1.1 Types of data (speech, text, images, gestures) and their relevance.
1.2 Refactoring and cleaning data for training interaction systems.
1.3 Evaluating data readiness and fitness for specific interaction use cases.
Practical Component
Demonstrations of preparing various types of data for an interaction AI system, identifying errors in data, and rectifying them.
Week 2 - Data Collection and Labelling
Learning OutcomeStudents will learn strategies for collecting data from different sources and ensuring the labels are accurate.
2.1 Collecting and labelling speech, image, and gesture data.
2.2 Identifying common labelling mistakes and their impact on AI outcomes.
2.3 Strategies for improving data labelling for better training results.
Practical Component
Practical exercises in labelling datasets and checking for errors, followed by improving labelling strategies based on real-world interaction scenarios.
Week 3 -Integrating Different Data Types
Learning OutcomeStudents will learn how to combine and integrate various forms of data for more powerful interaction AI systems.
3.1 Understanding the challenges of multi-modal data (text, speech, images).
3.2 Techniques for blending data sources to enhance interaction quality.
3.3 Ensuring the integrity and accuracy of integrated datasets
Practical Component
Students will integrate multi-modal data (e.g., combining speech data with text input) and assess the interaction AI systems effectiveness.
Week 4 -Data Quality and Selection for Interaction AI
Learning OutcomeStudents will gain the ability to evaluate and select the appropriate data to achieve accurate interaction AI results.
4.1 Criteria for selecting high-quality interaction data.
4.2 Identifying gaps in data and methods for data augmentation.
4.3 Evaluating the sufficiency of the data for various interaction AI tasks.
Practical Component
Practical exercises in assessing data quality, performing data augmentation, and selecting the right datasets for specific interaction AI applications.
Week 5 -Managing Insufficient Data
Learning OutcomeStudents will learn how to handle situations where data is insufficient and how to generate synthetic data for training.
5.1 Techniques for generating synthetic data using AI and ML tools.
5.2 Using transfer learning to overcome data limitations.
5.3 Assessing the sufficiency of data in terms of model performance.
Practical Component
Hands-on practice in generating synthetic data and applying transfer learning techniques to solve data scarcity issues.
Week 6- Project and Assessment
Learning OutcomeStudents will work on a custom data preparation project for a real-world interaction AI application, and present their findings.
6.1 Selecting a data collection strategy for an interaction AI system.
6.2 Peer assessment and evaluation of the projects effectiveness.
6.3 Final presentations and sharing insights from the project work.
Practical Component
Students will present their data management projects, demonstrating their ability to handle and prepare data for interaction AI systems