Code for Phonotactic mechanisms behind contact-induced morphological simplification in Alorese: An agent-based model of intergenerational transmission
This repository contains the code for the paper Phonotactic mechanisms behind contact-induced morphological simplification in Alorese: An agent-based model of intergenerational transmission by Peter Dekker, Marian Klamer and Bart de Boer. The code has been developed by Peter Dekker and conceptual contributions have been made by Marian Klamer and Bart de Boer.
The data used for the model (in folder data) is based on a list of prefixing verbs and prefix and suffix paradigms on pages 31-32 of: Nishiyama, K., & Kelen, H. (2007). A grammar of Lamaholot, Eastern Indonesia. Lincom Europa.
This script requires Python 3.9 or later. To install the dependencies use pip and the requirements.txt in this directory. e.g.
$ pip install -r requirements.txt
The evaluation.py script can be invoked to run the model from the command line and plot the results to files. There are two modes to run the script:
Use the --evaluate_prop_l2 mode to evaluate the effect of the proportion of L2 speakers on both prefix and suffix complexity. This mode has been used to generate most plots in the main article. Model settings can be set by supplying the setting name and one value, for example the probability of phonotactic reduction and the generalization probability in this case. In addition, meta-variables generations and iterations can be set. --runlabel is used to set a custom label for the output file names.
python3 evaluation.py --evaluate_prop_l2 --generations 100 --iterations 20 --interactions_per_generation 50 --phonotactic_reduction_prob 0.3 --generalization_prob 0.2 --runlabel phonred03
With the --evaluate_param mode, one can evaluate different values for a specific parameter, and plot the outcomes of the different values as lines in one plot. This is useful to evaluate different settings for for examle the lexical concept generalization and phonotactic reduction in one plot, as done in the SI of the article. To keep the plot clear, it only shows the suffix L2 complexity. Give the parameter you want to evaluate as option, in this case --phonotactic_reduction_prob and give a list of settings. No other model parameters can be given at the same time, but meta-parameters (generation, iterations) can be given.
python3 evaluation.py --evaluate_param --generations 100 --iterations 50 --phonotactic_reduction_prob 0.0 0.25 0.5 0.75 1.0 --runlabel evalphonred
config.py defines the user-settable parameters that can be given to the evaluation script as command line parameters. Some relevant parameters are:
--phonotactic_reduction_prob: set to 1.0 to turn phonotactic reduction on--phonotactic_reduction_drop_boundary_phoneme: set to False for full reduction of the affix, set to True for reduction of the boundary phoneme of the affix--generalization_prob: lexical concept generalisation, set to 1.0 to have full generalisation.
From the csv file from a previous model run, a plot can be generated, without rerunning the simulation, using the option --plot_from_raw followed by the path to the csv file. In addition to this option, the mode should still be given (evaluate_prop_l2 or evaluate_param), in order to create the right plot. For example, to evaluate proportion L2:
python3 evaluation.py --plot_from_raw output-oldplot/proportion_l2.csv --evaluate_prop_l2 --runlabel newplotlabel
Example to evaluate the parameter generalization_prob:
python3 evaluation.py --plot_from_raw output-oldplot/generalization_prob-evalparam.csv --evaluate_param --generalization_prob 0.0 0.01 0.05 0.1 0.5 1.0 --runlabel newplotlabel