WALS Online (Dryer, Haspelmath 2013) is a popular typological database, in which languages are described w 7F7A ith respect to a large selection of typological features. Each feature selects a set number of contrastive types (values), e.g. "No tone", "Simple tone system", "Complex tone system" for WALS feature 13A "Tone". Languages are then assigned one of these values for their respective features.
For some types of data analysis, however, it can be useful to look at the features differently. For instance, it might make sense to make a binary distinction for Tone, between "No tone" and "Tone system" instead of the ternary distinction defined in WALS. Such transformations are trivial to accomplish: in the example above, on only needs to recode all instances of ""Simple tone system" and "Complex tone system" as "Tone system".
This repository contains a list of useful WALS data transformations (see recode-patterns.csv), as
well as an R script (wals-recode.R) that generates a table containing both the original and
transformed WALS data.
The file recode-patterns.csv is a comma-separated table of recoding patterns which describes how
WALS data should be transformed. Every row in this table describes a single (transformed or
original) feature, which will appear as a column in the table created by the script. The relevant
table rows are:
-
wals.fname: WALS feature name, exactly as it appears in WALS Online data.
-
new.fname: New variable name, to appear in the generated data table.
-
wals.levels: List of feature values (one per line in the cell), exactly as they are defined and appear in WALS Online data. Every value is preceded by a unique integer ID that will be used in the recoding pattern. Different values are separated by a new line.
-
new.levels: List of feature values after transformation. These values will be present in the generated data table.
-
recode.pattern: Describes how feature values should be recoded. The pattern has the general form
A-B-C, with as many hyphen-separated components as there are new (post transformation) feature values. Every component describes how the respective (first, second etc.) feature value should be formed. A component can be either a single integer number — in which case the original feature value with that ID is taken — or a slash-separated list of original value IDs (e.g. 1/2/5) — in which case these values are collapsed together.
If all relevant columns are filled out for a row, the corresponding feature is copied from WALS data, transformed according to the pattern, and stored as a column in the generated table. If only wals.fname is filled out for a row, the corresponding feature is copied from WALS without transformations.
To run the transformations, simply execute the R script in R/wals-recode.R. The script expects
the working directory be set to the root of the repository. It expects to find WALS Online data in
the folder wals-dataset and the table of data transformations in the comma-separated file
recode-patterns.csv. The script will create a table of feature values per language in WALS and
save it in the file wals-recoded.csv.
which contains a table of languages (rows) and transformed WALS features (columns).
The script additionally executes a subsequent data cleaning step (fix-languages.R), which removes
some inconsistencies in how language data is treated (see that script for details).
The recoding are based on WALS Online data as published on http://wals.info on 2016-06-23. The scripts have not been tested with other versions of WALS Online data. Copy of WALS online from 2016-06-23 is included within this repository for convenience.
Data of WALS Online is published under the CC-BY 4.0 license: http://creativecommons.org/licenses/by/4.0/
Scripts and aggregation metadata in this repository is published under CC-BY 4.0 license: http://creativecommons.org/licenses/by/4.0/
It should be cited as
Bickel, Balthasar, Taras Zakharko. 2018. Recodings of WALS Online data. https://github.com/IVS-UZH/WALS-recodings
Dryer, Matthew S. & Haspelmath, Martin (eds.) 2013. The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info, Accessed on 2016-06-23.)