The Potential of Automatic Word Comparison for Historical Linguistics: Source Code and Data
This repository provides source code and data for the paper on the potential of automatic word comparison for historical linguistics (List, Greenhill, and Gray).
You will need LingPy (v >= 2.5, http://github.com/lingpy.lingpy), python-igraph (http://igraph.org/python-igraph), and the requirements which are needed for both packages. You will also need R if you want to make all plots that we prepared for the analyses.
To simply run all analyses, open your terminal, make sure you cd ino the folder and type:
$ make all
This will run both the training and the test analyses. Note, however, that this analyses uses the results of the Monte-Carlo permutation, which is very time-consuming, and whose results we submit along with the data-files which have a "bin.tsv" extension. If you want to re-run the Monte-Carlo permutation, you should first remove these files from the directory. You can also do this by typing:
$ make clean
If you run the analyses along with the Monte-Carlo permutation, this will take some time. All code pieces which you need to run the analyses are prefixed by
C_ in the file-collection.
All datasets in this sample are prefixed by
D_. There are 12 datasets, 6 training datasets, prefixed by
D_training, and 6 test datasets, prefixed by
D_test. For the sources of the data, please consult the paper, which is submitted along with this repository.
The Infomap plugin (see the pape for details) is in the file
P_infomap.py. As of lingpy-2.5, this plugin is also regularly integrated into the library. However, this repository provides the original version that uses it as a plugin to lingpy.
Output Created by the Scripts
There are four kinds of output created by the scripts:
- simple tsv-files, all prefixed with
training, depending on which file was analyzed,
- plots, all prefixed with
- LaTeX tables, all prefixed with
- txt-result files, all prefixed with
If you run into problems or have further questions regarding the data, please contact us writing an email to email@example.com.