Run main_swda_elmo_mean.py for training (set train = True)
the models, else it will return the context and non-context models
with direct function predict_classes_elmo_mean_features.
Similarly, main_predictor_online.py is for models of
ELMo normal features.
They are already called in the following scripts for their usage,
and hence before running following scripts,
directory params should contain all the weight files and
directory features should contain all the elmo features.
-
To annotate IEMOCAP data: run
mocap_dia_act_annotate.pyto get context and non-context annotations and save as numpy arrays, and generate final annotation (csv) file. -
To annotate MELD data: run
meld_dia_act_annotate.pyto get context and non-context annotations and save as numpy arrays, and generate final annotation (csv) file.
-
Run the ELMO embedding server: running
elmo_emb_server.pywill start the embedding server. -
Run DA annotator server: running
dia_act_annotator.pywill start the annotator server. -
To annotate your data: run
example_api.pyscript to annotate your own utterances.