File struture explained below:

DEPLOYMENT

main_pipeline_whisper.py -> runs prediction models based including whisper transcription main_pipeline.py -> runs text-based-input prediction (was used during development) user_interface.py -> localhost web interface to deploy the above main_pipeline including whisper

TRAINING

goemotions_training_template.py goemotions_training_trans_1.py goemotions_training_conv2.py goemotions_training_conv1.py

imdb_training_template.py imdb_training_gru.py imdb_training_conv1.py imdb_training_conv2.py

glove.6B.50d.txt -> pretrained embedding layer used for the entire project

TESTING

TESTING THE FULL PIPELINE INCLUDING WHISPER

whisper_test.py -> creation of test transcription, which are manually classified and flagged if mistake happend test_full_pipeline.py -> loads best performing sentiment model and predicts sentiments for the test transcriptions sentiment_analysis_metrics.py -> creates metrics for full pipeline sentiment_visualization.py -> creates plot for full pipeline metrics

TESTING SENTIMENTS

imdb_llm_test.py -> uses the api to predict sentiments on the test set imdb_model_comparison.py -> computes and compares accuracy measures of sentiment models (including trained models and LLM)

TESTING EMOTIONS

load_test_goemotios.py -> creates test dataset used for LLM and model comparison goemotions_llm_test.py -> uses API call to predict emotions on test set & computes metrics for comparison goemotions_model_comparison.py -> computes and compares accuracy measures of emotion models (including trained models and LLM)

API CONNECTION

llm_api.py -> sets up function to use API call to predict tuple of (sentiment, emotion)

FUNCTION HOLDER

prediction_pipeline.py -> holds one prediction function used for code clarity