Working with Digital Data
in Religious Studies
10. Advanced Processing & AI: Get a Grip on Big Data
Summer Semester 2024
Prof. Dr. Nathan Gibson
Outline
- Review: Metadata & FAIR Data
-
Principles of AI/Machine Learning
–Break–
- Playtime!
- Critical reflection
Project & Presentation
5 minutes:
- What is the format and source of your data? (Include critical reflection.)
- How did you edit, process, filter, or add to it? Why? (Show how you have used at least one of the approaches we discussed in class.)
- What is the most interesting question you might answer with your dataset?
- What is the most valuable thing you learned in the process?
A few slides or screen-sharing is allowed, but make sure you can keep it to 5 minutes!
Exams & Term Papers
Sign up by next Friday!
Term paper: Set up a meeting with me if you haven’t already.
1. 📈 Metadata & FAIR Data Review: Learning Objective
Assess whether and how to make your data more open.
1. 📈 Metadata & FAIR Data Review: Metadata
Metadata: data about your data
- research metadata: creators, process, guidelines, version information
- decisions about metadata should reflect data format and research purpose
- metadata lets people find, connect to, and keep track of your data
1. 📈 Metadata & FAIR Data Review: FAIR Principles
F
A
I
R
1. 📈 Metadata & FAIR Data Review: Findable
unique identifiers, metadata, in a searchable resource
1. 📈 Metadata & FAIR Data Review: Accessible
data can be accessed using a standard system on the basis of identifiers
1. 📈 Metadata & FAIR Data Review: Interoperable
data is in a format that can be used by common systems and is linked to other datasets
1. 📈 Metadata & FAIR Data Review: Re-usable
data is licensed for re-use, source is known, meets community standards
1. 📈 Metadata & FAIR Data Review: Should you make your data FAIR?
🧭 Today’s Learning Objective
Critically explore the relationships between the inputs and outputs of machine-learning models (artificial intelligence).
2. Principles of AI/Machine Learning: Big Data
Big Data: Data that defies “traditional methods” of processing or analysis because of its large scale.
2. Principles of AI/Machine Learning: Big Data
Examples:
- The Facebook Graph: ca. 3 billion users (and the relationships between them!)
- GPT-3 training data: 500 billion words
2. Principles of AI/Machine Learning: Big Data
Humanities examples:
- Thousands of handwritten manuscripts (too many to transcribe and collate manually)
- The Louvre art collection (too many for traditional art criticism)
- Social media posts about “pagans” (too many languages and search terms)
2. Principles of AI/Machine Learning: Artificial Intelligence
Artificial Intelligence (AI): a vague term used for
- science-fiction computers that take over the world (HAL 9000, etc.)
- chat bots: things that “communicate” in a human-like way
- generative AI: bots that create text, images, music, software, etc.
- machine-learning algorithms and models
2. Principles of AI/Machine Learning: Machine Learning
Machine Learning: A process of using data to train software to recognize or predict patterns in new data
2. Principles of AI/Machine Learning: Machine Learning
What would happen if … ?
2. Principles of AI/Machine Learning: Machine Learning
Ground truth: Correctly labeled data used for training and testing
Neural networks use a process that turns nodes on or off based on many different inputs, and then goes back and refines the “weight” of these inputs.
Large language models predict the next word(s) after having been trained on a very large dataset.