somerandomguyontheweb

Not the maintainer, but I think you can just use the TFHub models available off the shelf, e.g.: ``` from easybert import Bert bert = Bert("https://tfhub.dev/google/bert_cased_L-24_H-1024_A-16/1") # large cased model...

Since you're providing an example in the prompt (proverb beginning + proverb ending), it appears that your eval might be restructured as a few-shot eval. See e.g. `sarcasm` or `anagrams`...

@kmike Hi, sorry for upping this old issue – is it still relevant to add support for Belarusian? If yes, please let me know, I'm interested in working on this.

As advised by Ukrainian colleagues who reviewed several hundred sentences in the sample, I'm adding more patterns to the rules file `uk.toml`. The number of extracted sentences decreased to 620'015,...

Some more sentences have been reviewed by native speakers of Ukrainian, and it is clear that the ratio of errors is still higher than acceptable. I'm putting this effort on...