WWW
As a researcher who wants to train a machine-translation type seq2seq model for hypothesis correction I would like to extract pairs of actual and predicted strings from speakers in the TORGO dataset.
A PR made to the torgo_testing repo with a script that is able to generate, actual and predicted string pairs for each speaker after running inference on our trained models
WWW As a researcher who wants to train a machine-translation type seq2seq model for hypothesis correction I would like to extract pairs of actual and predicted strings from speakers in the TORGO dataset.
Documentation of plan: https://docs.google.com/document/d/1aAzRwka9W3B8xB6kuggZy0P93VlbCZfjxQraok2Vl10/edit?usp=sharing
AC