Open mattroddy opened 5 years ago
Similar observation. Additionally it seems that the -p flag actually refers to the previous dialogue act and not the sentence as indicated in the comments (https://github.com/ColingPaper2018/DialogueAct-Tagger/blob/master/scripts/predict.py#L25) https://github.com/ColingPaper2018/DialogueAct-Tagger/blob/master/predictors/svm_predictor.py#L32
https://github.com/ColingPaper2018/DialogueAct-Tagger/blob/master/trainers/svm_trainer.py#L54
Thanks for the very useful tool. In line 40 of predict.py you aren't passing the previous sentence in so (unless I missed something) I don't think the model's predictions are changed by the -p flag.