Code and Data for the Paper "Comparing Styles across Languages"

Our holistic politeness dataset is free for non-commercial use and licensed under the CC-BY-NC license.

The repository is structured as follows:

├── MLC 
│   ├── original + expanded PoliteLex
│   ├── code for expansion and purification
├── analysis
│   ├── extracted shapley values
│   ├── code for feature set aggregation
├── data
│   ├── labeled holistic politeness dataset (train/test/val files for English, Spanish, Japanese, and Chinese)
├── models
│   ├── huggingface fine-tuning code