Second version of the dataset generated by my pipeline. Consists of 654 examples, about 2.3 million characters of text.
Improved over v1 by greater volume, reduced noise (headers, titles, copyright notices, contact information, graph labels, etc) and by pruning examples that have empty prompts or responses.