Skip to content

Latest commit

 

History

History
26 lines (17 loc) · 1.53 KB

README.md

File metadata and controls

26 lines (17 loc) · 1.53 KB
Raw

Measuring the Global Utility of Language Technologies

This is supplementary data and code to help reproduce the results of a paper of the same name. It is currently available in somewhat unpolished form, but will be fully documented in the near future.

File Structure

  • populations: contains .tsv files with language-population data and other population-relevant data

  • task_results: contains .tsv files with the aggregated results for each NLP task.

  • economic_indicators_data: contains files from WITS (under wits_en_trade_summary_allcountries_allyears) and converted to map to languages instead of countries. Important files are:

    • languages_to_gdp.tsv for monolingual mapping of languages to associated GDP estimations.
    • bilingual_indicators.tsv for bilingual mapping of languages to associated bilingual indicators (Imports, Exports) estimations. Also includes the triangulated BLEU scores for the language pair.

  • figs: contains correlation figures, created with the plot_*_correlations.py scripts.

  • area-classifier: contains data and code for a classifier of areas

Other files:

  • counterfactuals.py: computes the counterfactual scenarios presented in the paper
  • constants.py contains functions to read in all necessary data, which are used in other files to run the metrics estimations and produce the plots
  • economic_indicators.py: contains function to read in economic indicators (called by constants.py)

TO-DO

  1. Add general metric calculation script
  2. Data paths are all absolute, need to correct this