"The plural interpretability of German linking elements"

Roland Schäfer¹ & Elizabeth Pankratz² (accepted w/minor rev. on 3 June 2018 by Morphology [JOMO])

¹Freie Universität Berlin, ²Humboldt Universität zu Berlin

• Paper website http://rolandschaefer.net/?p=1532
• Data storage https://github.com/rsling/linkingelements
• Contact email mail@rolandschaefer.net

This is the README file for data package and paper LaTeX/knitr sources.

1. To replicate the full analysis, you have to get some large data files (if you checked this out from Github). To download them, type the following in Bash (MacOS or GNU/Linux - or hypothetically Windows with Bash and POSIX tools) after descending to this folder:

   Data/get_bigfiles.sh

2. You can check how the count files and other data bases were generated by checking out the following Bash scripts:

   Data/Database/make_counts.sh
   Data/Database/make_data.sh
   Data/Database/make_real_blacklists.sh

This is usually not required, however. The generated files are included in the distribution, and it takes a long time to re-generate them.

3. To see how we did the data generation for the corpus study and creation of stimuli, check the following file:

   Data/R/corpus.data.R

4. To check how we analysed the (manually annotated) corpus data in R, see this file:

   Data/R/corpus.analyse.R

There is a large number of original query results (also downloaded by Data/get_bigfiles.sh, see 1) in

   Data/Corpusstudy/Queries/Output/

5. To check how we analysed the split-100 experiment in R, see this file:

   Data/R/split100.analyse.R

6. The data from the experiment (including PsychPy files), which are analysed by the script in step 5 is located in

   Data/Split100/

7. To see how we integrated the data into the main paper, see the knitr sources including the make file:

   Paper/leglossa.Rnw
   Paper/Makefile

Notice that it was not feasible to include the whole data generation and analysis process in the knitr file. Compiling the document would take hours.