Taxonomy creation

Tim: Let's have a conversation about using ChatGPT to help create a taxonomy for a particular knowledge domain. Keep your answers to less than a paragraph and conversational and finish each answer with a question back to me.

ChatGPT: Sure, I'd love to help you create a taxonomy! What specific knowledge domain are you looking to develop a taxonomy for?

Tim: First let's discuss taxonomies in general. Can you define a taxonomy and explain how it can be useful for professionals and specialists.

ChatGPT: A taxonomy is a structured classification system that organizes information into categories and subcategories based on shared characteristics. It helps professionals and specialists by providing a clear framework for organizing and accessing knowledge, making it easier to find, understand, and use relevant information. How do you think a well-defined taxonomy could benefit your specific field?

Tim: I have used taxonomies before in regulatory intelligence, software development, product management and healthcare. They provide a common understanding of the domain and allow for easier integration between different systems and vendors. They are also useful to help with training and education to help map out learning programs and training resources. What are the key considerations when trying to create a taxonomy from scratch?

ChatGPT: When creating a taxonomy from scratch, key considerations include understanding the scope and purpose of the taxonomy, involving domain experts, ensuring categories are mutually exclusive and collectively exhaustive, and making it user-friendly and scalable. It’s also important to regularly review and update the taxonomy to keep it relevant. Which domain are you currently focusing on for creating this new taxonomy?

Tim: I am interested by way of an example to create a taxonomy covering the entire field of Artificial Intelligence. Could you create a draft taxonomy for this knowledge area?

A number of prompts and responses were used to create a taxonomy for Artificial Intelligence. There was some tweaking and de-duplication and I then progressively asked ChatGPT to generate descriptions for each category and sub-category.

The result was a comprehensive 3-level taxonomy which has been published on the Explain Yourself site here. The taxonomy is also available as an XML file below.

And as an Excel file.

You are free to use this taxonomy in your own projects including:

1. Planning AI-related content and coming up with new ideas for articles, reports and research.
2. Classification of existing data, content, work items or projects.
3. Extension or relationship with other taxonomies.

Classification Example

As an example, I asked ChatGPT to use the taxonomy to classify the previous article published to Explain Yourself which was about music creation. This article is available here.

ChatGPT provided the primary classification

Category: AI Applications

1. Subcategory: Music and Art Generation
   • Item: AI-Generated Music: The article discusses the use of AI in generating music using an application called AIVA.

Which was clearly correct.

ChatGPT was also asked to provide secondary classifications. The results were as follows:

Category: Machine Learning

1. Subcategory: Deep Learning
   • Item: Generative Adversarial Networks (GANs): The article describes how AI generates music, which could involve GANs or similar deep learning techniques for creating new compositions.
   • Item: Neural Networks: The use of AI for music composition likely involves neural networks, which are a fundamental part of many AI generative models.

Category: AI Tools and Platforms

1. Subcategory: Frameworks and Libraries
   • Item: Music Generation Libraries/Tools: While not explicitly mentioned, AIVA could be seen as part of the broader category of tools and libraries used for music generation.
2. Subcategory: Cloud AI Services
   • Item: AI Music Generation Services: If AIVA or similar tools are cloud-based, this could fit within the scope of cloud AI services.

Category: AI Research and Development

1. Subcategory: Current Trends
   • Item: Generative AI: The article fits well within the context of current trends in generative AI, which includes creating music.

All secondary classifications are fair and reasonable enough. It is particularly useful to see the reasons why the classifications were made so that they could be reviewed and approved or rejected as required.

In Summary

ChatGPT is particularly good at creating taxonomies, describing them and generating different versions for both human and machine consumption. Taxonomies can be used in a variety of ways including:

• Mapping a specific knowledge area, domain, product or industry.
• Providing a common standard to be used across different platforms and products.
• Planning educational resources and curriculae.
• Generating ideas for marketing content.