mahowak / deictic_adverbs Goto Github PK

• figures: all the figures used in the manuscript (Figures 1-8)

• sheets: all data files generated by the scripts

• readable_data_tables: all the spatial deictic demonstratives complied by Nintemann et al. (2020), as spreadsheets

• deictic_appendix_docs: the original appendix in Nintemann et al. (2020)

• Fit_gamma.ipynb: the main file. It executes run_ib.py to perform the relevent calculation and generates the data files.

• run_ib.py: the Python script for calculating the informativity and complexity of real lexicons and all possible lexicons, given a number of distal levels, decay parameter $\mu$, and place/goal/source coordinates

• ib.py: the Python script for the Information Bottleneck (IB) method in general

• stirling.py: the Python script for calculating the Stirling number of the second kind, given 2 parameters $(n, k)$. Code credit to EXTR3ME.

• get_prior.py: the Python script for interpolating the prior / need probability $p(m)$ from the Finnish corpus

• get_lang_data.py: the Python script for processing the data from Nintemann et al. (2020)

• enumerate_lexicons.py: the Python script for generating all possible lexicons

• systematicity_combined.py: calculate the systematicity score

• plot_figs.R: the R script for generating the main figures (Experiments 1 and 3)

• plot_grid_search.R: the R script for analyzing the grid search data (Experiment 2)

1. Launch Fit_gamma.ipynb. Parameters in the model are:

• logsp: the trade-off parameter array. The default setting np.logspace(0,2, num = 1500) generates 1500 log-spaced tradeoff parameters $\beta$ from 1 to 100.

• mu: the decay parameter. The default is 0.3.

• num_dists: the number of distal levels. The default is 3.

• pgs_dists: the coordinates of place/goal/source. The default is [0, 0.789, -1.315]

  The script will generate 4 files, saved in the sheets folder:

• 2 data files for real lexicons and all possible lexicons, respectively.

  • File 1: a data file for real lexicons, syntax: real_lexicons_fit_mu_<mu value>_pgs_<pgs values>num_dists_<number of distal levels>.csv (the default name is real_lexicons_fit_mu_0.3_pgs_0_0.789_-1.315num_dists_3.csv)
  • File 2: a data file for all possible lexicons, syntax: sim_lexicons_fit_mu_<mu value>_pgs_<pgs values>num_dists_<number of distal levels>.csv (the default name is sim_lexicons_fit_mu_0.3_pgs_0_0.789_-1.315num_dists_3.csv)

• 2 files for the deterministic and non-deterministic optimal frontier, respectively.

  • 1 data file for the deterministic optimal frontier, syntax: ib_curve_deter_mu_<mu value>_pgs_<pgs values>num_dists_<number of distal levels>.csv
  • 1 data file for the non-deterministic optimal frontier, syntax: ib_curve_non_deter_mu_<mu value>_pgs_<pgs values>num_dists_<number of distal levels>.csv

2. Open a terminal on the repo, run the lines below, assuming you're following the default condition

python systematicity_combined.py --filename sheets/real_lexicons_fit_mu_0.3_pgs_0_0.789_-1.315num_dists_3.csv

python systematicity_combined.py --filename sheets/sim_lexicons_fit_mu_0.3_pgs_0_0.789_-1.315num_dists_3.csv

2 new files will be generated, with the systematicity scores attached.

3. Open and run the plot_figs.R script. Figures will be generated in the figures folder.

1. Open a terminal on the repo, run the lines below, assuming you're following the default condition.
   • Place/goal/source coordinate search: Run the code below (warning: this might take a day to execute. We recommend skipping this step and using our output file.). This will generate a file called total_grid_search_gridsearch.csv in the sheets folder.

python run_ib.py --total_search --outfile sheets/total_grid_search

- Prior search: Run the code below. This will generate a file called ``prior_search_gridsearch.csv`` in the ``sheets`` folder.

python run_ib.py --prior_search --outfile sheets/prior_search

- Mu search: Run the code below. This will generate a file called ``mu_search_gridsearch.csv`` in the ``sheets`` folder.

python run_ib.py --mu_search --outfile sheets/mu_search

2. Open and run the plot_grid_search.R script. Figures will be generated in the figures folder.