Kenneth Enevoldsen

Hi @jasonwilliams I have an a similar issue: I have a case here: ``` # Topic header Description of topic, notes etc. first card: ## card 1 front1 % back...

Stopping on H1 would help in my case. But adding an optional stopping token using regex would also be an option. Another option would be to change the regex for...

It seems it might be relevant to also remove ignore comment i.e. ``.

Sure. However, do believe that a field with a regex for extracting the card whether it be HTML comments or using another syntax would be ideal. This would allow the...

I Think HTML comment is a poor default though as it is also a comment thus I would expect it to ignore it in markdown documents potentially one of these...

Indeed all the resources are easily available. This is simply redistributing an application of a trained model which could be easily reproduced. Thus some kind of hub for projects would...

Hi @adrianeboyd. The BERT by botxo is a good option to go with. Alternatively the Danish ÆLÆCTRA (which is a small-sized Electra also uploaded by @Maltehb, I believe it is...

Thanks for this suggestion @martincjespersen. Def. a good suggestion.

This would require solving issue #14 as the model currently takes in one example at a time, which would be unfeasible for back-translation.

- pseudo-perplexity for MLM models like BERT. - entropy Potentially see [this](https://thegradient.pub/understanding-evaluation-metrics-for-language-models/) blog post.