This repository provides code to replicate the analyses in the following
work:
Ramezani, A., Stellar, J.E., Feinberg, M., and Xu, Y. Evolution of the moral lexicon, Open Mind (2024).
To run scripts and notebooks, modify `__dir__` accordingly.
> Requirements
python >= 3.8
jupyter
> Python packages
numpy
pandas
en_core_web_sm
changepoint
scipy
gensim
scikit-learn
statsmodel
seaborn
> Data
1. HistWords diachronic embeddings:
https://nlp.stanford.edu/projects/histwords/
2. Moral Foundations Dictionary version 2: https://osf.io/whjt2/
3. Historical Thesaurus of English: https://ht.ac.uk
4. Concreteness ratings: http://crr.ugent.be/archives/1330
5. Valence ratings: http://crr.ugent.be/archives/1003
6. Metaphor Map of English: https://mappingmetaphor.arts.gla.ac.uk
7. The Intercontinental Dictionary Series: https://ids.clld.org
> Scripts
Clone change point detection algorithm from https://github.com/viveksck/changepoint and put it next to scrpits.
Code to reproduce the results.
`python scripts/moral_sentiment_change.py`
`python scripts/word_grouping.py`
`python scripts/detecting_change_points.py`
`python scripts/concreteness_valence_timeseries.py`
`python scripts/domain_shift.py`
> Notebooks
After replicating the data, run the following notebooks to reproduce the paper experiments and illustrations:
`MFD_concreteness_results.ipynb`: (Study 1) Corpus-based evidence for concrete-to-abstract shift in MFD words
`MFD_valence_results.ipynb`: (Study 1- appendix) Corpus-based evidence for valence shift in MFD words
`MFD Concreteness Valence.ipynb`: (Study 1- appendix) Experiment on concreteness-valence dependency.
`HTE valence ratings.ipynb`: Extarcting HTE valence ratings.
`HTE analysis.ipynb`: (Study 1) Corpus-based evidence for concrete-to-abstract and valence shift in HTE words.
`domain shift.ipynb`: (Study 2) Cross-domain word meaning change
`MME.ipynb`: (Study 3) Historical evidence for abstraction in lexicalized metaphors of moral words
`compound_analysis.ipynb`: (Study 4) Metaphorization for compound moral words
