Open dchest opened 1 month ago
Hi @dchest
We appreciate your contribution in explaining the issue with relevant examples. You are right about collision with some other sentence structures which might not yield an accurate sentence breakdown.
In our current eng-lite model, Markdown format training hasn't been included. Adding it to our list of enhancements.
Shall keep you posted once the MD trained model is ready.
Many thanks, Rachna
In plain text and Markdown it's common to use lists like this:
Unfortunately, the model doesn't consider such list items as separate sentences. Is there a possibility for improvement here? For example, consider a single line break as an indicator that the sentence could end? I assume there's a collision with some other sentence structure that makes it necessary to consider those as a single sentence?
Sentences:
Two line breaks, however, make it work:
Sentences:
Thank you!