Kenneth Heafield

Results 290 comments of Kenneth Heafield

Yeah in general the MT system would need to be trained with a placeholder to do this correctly.

In general there can be an OOV like this in the middle of a sentence. The langid would be as correct as it can be. If there is an OOV...

I put out a call for MSc student projects at the university, letting them pick which platform including LibreOffice. Looks like one will be doing Android app and two will...

On a tangential note, https://github.com/jerinphilip/lemonade did an ibus plugin with a demo in libreoffice: https://www.youtube.com/watch?v=Lsmpx_A_7Y8

We don't use faiss currently. It could be in the future if we switch to maximum dot product instead of shortlists.

Hi, welcome! Have you taken a look at https://github.com/browsermt/bergamot-translator/pull/437 ? @jelmervdl is the javascript expert, I defer to him but it does look much simpler.

@jerinphilip Your short-term goal is to provide an implementation of Abhishek's C API using ruy so that it can be integrated into Gecko (part of Firefox) for ARM support. This...

"Attention is not not Explanation" https://aclanthology.org/D19-1002.pdf "Attention is not Explanation" https://aclanthology.org/N19-1357.pdf

The alignments come from guided alignment trained from fastalign. Not from attention. The alignments are what drives HTML alignment.

Actually this may not be a bergamot-translator issue at all. We should just build the normalization rules into the SentencePiece vocab. https://github.com/marian-nmt/marian-examples/tree/master/training-basics-sentencepiece