search

declare-lab / MELD

MELD: A Multimodal Multi-Party Dataset for Emotion Recognition in Conversation
GNU General Public License v3.0
788 stars 200 forks source link

The meanings about the features. #4

Closed XinyuLyu closed 5 years ago

XinyuLyu commented 5 years ago

image Can you exlpain what features each file represents？I'm a little confused about their file names. After reading the codes in basline.py and data_helper.py, you use cnn to extract the textual-features. Does it mean the vedio features? In your fusion model, I can't find the vedio branch. And what does text_glove_average_emotion.pkl mean? And what's the difference between audio_embeddings_feature_selection_emotion.pkl and audio_emotion.pkl?

soujanyaporia commented 5 years ago

text_glove_average_emotion.pkl -> initialise words with glove and then take average to represent a sentence video -> we do not provide video features audio_embeddings_feature_selection_emotion.pkl -> audio features after doing feature selection using PCA