changesubjectmarkers

Background

This code accompanies the following paper (Evolang 2024 presentation & proceedings paper), which describes the analysis performed using this code:

Peter Dekker, Sonja Gipper & Bart de Boer (2024). 3SG is the most conservative subject marker across languages: An exploratory study of rate of change.

A cross-linguistic analysis of the rate of change of different subject markers is performed, by comparing proto and modern forms from the following dataset:

Ilja Seržant. (2021). Dataset for the paper "Universal attractors in language evolution provide evidence for the kinds of efficiency pressures involved" [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7641119

Installation

With the script analyse_data.py, an analysis of the rate of change of person markers for different grammatical persons is performed.

The script requires a few installation steps:

1. The script uses R, so install R, CMake and some dependencies needed for the R packages beforehand. On an Ubuntu machine:

sudo apt install r-base cmake libharfbuzz-dev libfribidi-dev

2. Install the Python package requirements: pip3 install -r requirements.txt (or with sudo for global install)

3. Install the required R packages: python3 install_r_pkgs.py (or with sudo for global install)

4. This script requires the file verbal_person-number_indexes_merged.csv from https://zenodo.org/records/7641119 (version v5) to be in the folder data.

5. Now you can run the script by running: python analyse_data.py. It will output plots in the output_data directory. For the model using unnormalised Levenshtein distance, set the variable NORMALISATION to none. For the normalised model, where the Levenshtein distance is divided by the length of the longest form, set NORMALISATION to max.