Description of the pipeline

If you want to use the code developed for the cultural analytics paper, then you would need to follow the steps below:

• To train the embeddings model, use the Java code developed by David Bamman in this repository. For the cultural analytics paper, we modified this Java code to emit output embeddings and include an additional facet. This modification is available in the form of a pull request (as of 3/31/2022) here.

If you're interested in the processed data, i.e., the embeddings from the data used in the paper, it can all be downloaded from this link

• Next, semantic changes are learned by running pipeline_temporal.sh
• Next, the leadership stats are calculated by running pipeline_sources.sh
• Finally, the randomization experiments are calculated by running pipeline_randomization.sh

Cite

If you use the processed data (embeddings from abolitionist newspaper corpus), please consider citing our data link as:

@data{DVN/EWYMFG_2021,
author = {Soni, Sandeep and Klein, Lauren and Eisenstein, Jacob},
publisher = {Harvard Dataverse},
title = {{Abolitionist Networks: Modeling Language Change in Nineteenth-Century Activist Newspapers}},
year = {2021},
version = {V1},
doi = {10.7910/DVN/EWYMFG},
url = {https://doi.org/10.7910/DVN/EWYMFG}
}

If you end up using the code from this repository, please also consider citing our paper as:

@article{soni2021abolitionist,
  title={Abolitionist Networks: Modeling Language Change in Nineteenth-Century Activist Newspapers},
  author={Soni, Sandeep and Klein, Lauren F and Eisenstein, Jacob},
  journal={Journal of Cultural Analytics},
  volume={6},
  number={1},
  pages={18841},
  year={2021},
  publisher={Department of Languages, Literatures, and Cultures}
}

Contact

Please contact Sandeep Soni (soni.sandeepb@gmail.com) or Lauren Klein (lklein@gmail.com) for any inquiries about the data or code.