Adjust the pipeline to try out different kinds of queries that provide better, i.e., more useful and meaningful, summaries of items of business. Maybe work with different branches? Make sure the different test branches use the same sample of business items.
Current ChatGPT query does not provide satisfactory results. https://github.com/fabianaiolfi/Parliamentarians/blob/5b549cb68a525e8907a9b1c1a4b22888af12ab8e/scripts/02_clean_data.R#L34
Adjust the pipeline to try out different kinds of queries that provide better, i.e., more useful and meaningful, summaries of items of business. Maybe work with different branches? Make sure the different test branches use the same sample of business items.