Content Meeting Session – 3rd June 2024

🌐 Discussion Topic:
Using Token Classification to Study Historical Attitudes Towards Plants and Animals

This event focused on an innovative approach to mining and analyzing Dutch attitudes towards plants and animals from 1550 to 2000, utilizing computational humanities and natural language processing (NLP).

🔍 Overview of the Presentation
This approach integrates cultural history, literary studies, and computational methods to track societal views on living nature. The traditional approach involves annotating data and training Named Entity Recognition (NER) models, but it explores using AI to streamline and enhance this process through the Large Language Models as Active Annotators (LLMaAA) method.

🚨 Highlights of Insights

• LLMaAA Principle: Establish clear definitions, manually label test and demonstration sets, select and annotate unlabeled examples with a GPT model, and train an LLM model on the annotated data.
• Current Progress: Comparing GPT annotations with human annotations.
• Future Goals: Demonstrate the effectiveness of LLMaAA for historical NER tasks, aiming for high precision, accuracy, recall, and cost-efficiency.

🌟 Presenter
This insightful session was presented by Arjan van Dalfsen.