Python Scripts for extract texture features given transcripts.

The extractor BERT is from this the repository: sentence-transformer;

the pretrained weights are from: distiluse-base-multilingual-cased-v2 according to the paper Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation. Advantages of this model according to the paper: Aligned feature space. 1) Vector spaces are aligned across languages, i.e., identical sentences in different languages are close; 2) vector space properties in the original source language from the teacher model M are adopted and transferred to other languages.

• python: 3.6+
• Pytorch: 1.7+

1. download training_data_transcripts in the root directory

training_data_transcript/
├── animals_transcripts1_train
    ├── 
    ├── 
    ├── 025157
        |── 025157_animals.srt
    ...

2. use process_srt_files in preprocess.py: to generate the raw data raw_data.npy that contains the list of chunks. Each chunk contains text, duration, talk_type, participant_id. One can use/modify ChunksDataset to load the raw data.

3. just run feature_extraction.py to extract the features and obtain the embeddings.npz. Each item contains a feature embedding and a participant label. One can use/modify EmbsDataset to load the extracted features.

[email protected]