Multimodal Sentiment Analysis Experimental Data (Processed MOSI/MOSEI Subset)

It contains aligned and processed features and intermediate artifacts derived from the public dataset(s) <e.g., CMU-MOSI / CMU-MOSEI>. Original videos are NOT included; only processed text/audio/vision features and labels are provided.[Contents & Structure]- data/ - train/ valid/ test/ (same split as the original or our reproduction) - modalities/: {text.csv, audio.csv, vision.csv} - labels/: {sentiment.csv or emotion.csv}- code-snippets/: loading examples and preprocessing scripts- README.md: field definitions and usage[Formats & Size]- Text features: <e.g., GloVe 300d or BERT-base pooled>, CSV/NPZ- Audio features: <e.g., COVAREP/opensmile, sampling and dims>- Visual features: <e.g., OpenFace/Facet; frame-level to segment aggregation>- Total size: <xx GB>[Labels & Splits]- Sentiment: <range or discrete classes>- Emotions (if any): <N classes>- Splits follow the original dataset or our scripts for reproducibility.[How to Use]We provide Python examples and EMOE-compatible loaders (works with PyTorch-DirectML/CPU).[Citation]Please cite <Your Name/Institution> and the original dataset paper(s) (e.g., CMU-MOSI/MOSEI).

Copyright and terms of the original data remain with their owners; this release contains derived features only for research reproducibility.