The History of Text Generation

Yes in the sense that there are libraries that can generate text (e.g. SimpleNLG), but you need to write code that does the analysis and chooses the content to show.

The "modern" way would be to train a seq2seq model. But for that you need training data, and it's harder to control the correctness of the results.

For a start, you can look at the slides from this summer school https://nlgsummer.github.io and read up on the BabyTalk project (http://www.nlg-wiki.org/systems/BabyTalk). SimpleNLG is indeed simple to use, but maybe for your task a template-based system (e.g. https://github.com/TallChris91/PASS) is enough.

TLDR: you either need to write code for managing the time series, or need a large enough dataset to train on. Or rely on commercial services such as those offered by Arria

Although my general rule is that if Javascript is required to read a site's content, it's probably not worth it. With the storied NoScript being released for Chrome (https://news.ycombinator.com/item?id=19650731), I suspect I'll have more and more company as the new web gets more abusive and dangerous.