CorporaCoCo

The package implements the method introduced in Wiegand and Hennessey et al. (2017a). It identifies significant difference in co-occurrence counts for a given node or set of nodes across two corpora, using a Fisher’s Exact test.

A good place to start is the ‘Introduction to CorporaCoCo’ vignette. You can open the vignette with vignette("intro", package = "CorporaCoCo"). For a list of all documentation use library(help="CorporaCoCo"). For updates on development versions of the package and documentation, please watch this GitHub page.

References

• Wiegand, V., Hennessey, A., Tench, C. R., & Mahlberg, M. (2017a, May 24). Comparing co-occurrences between corpora. 38th ICAME conference, Charles University, Prague.

• Wiegand, V., Hennessey, A., Tench, C. R., & Mahlberg, M. (2017b, July 24). A cookbook of co-occurrence comparison techniques and how they relate to the subtleties in your research question. 9th International Corpus Linguistics Conference, University of Birmingham, Birmingham.

A very simple example of usage

This example takes the two Dickens novels 'Great Expectations' and 'A Tale of Two Cities' and compares the co-occurrences of a set of body part nouns. The idea is that since body part nouns are common in suspensions the statistically significant co-occurrence differences should include personal pronouns reflecting the differing narrative voices of the texts. We use the texts here only as sample data; we retrieve them from the CLiC API using the clicclient package (the other functions of that package are still under development; please see the clicclient GitHub page for details).

#devtools::install_github("mahlberg-lab/clicclient")
library(clicclient)

# retrieve texts for 'A Tale of two Cities' (TTC) and 'Great Expectations' from the CLiC corpora using the 'clicclient' package
TTC <- clic_texts("TTC")
GE <- clic_texts("GE")

# tokenize / create corp_text objects
TTC_text <- corp_text(TTC)
GE_text <- corp_text(GE)

# count co-occurrences / create corp_surface objects
TTC_cooccurs <- corp_surface(TTC_text, span = "5LR")
GE_cooccurs <- corp_surface(GE_text, span = "5LR")

 # set the body part nodes
nodes <- c('back', 'eye', 'eyes', 'forehead', 'hand', 'hands', 'head', 'shoulder')

# run co-occurrence comparison with corp_coco
results <- corp_coco(TTC_cooccurs, GE_cooccurs, nodes = nodes, fdr = 0.01)

results

##          x   y H_A  M_A H_B  M_B effect_size  CI_lower   CI_upper      p_value   p_adjusted
##   1:  back  me   3 1347  49 2391   3.2014513  1.565327  5.5296252 5.771195e-07 5.638457e-04
##   2:  back  my   1 1349  31 2409   4.1171190  1.528647  9.4632893 1.931898e-05 9.437321e-03
##   3:  eyes   i  10 1640  53 1747   2.3142117  1.318110  3.4597915 1.320717e-07 6.114918e-05
##   4:  eyes joe   0 1650  16 1784         Inf  1.831254        Inf 3.641000e-05 7.000398e-03
##   5:  eyes  me   3 1647  25 1775   2.9502767  1.233703  5.3219224 3.779912e-05 7.000398e-03
##   6:  eyes  my   5 1645  58 1742   3.4522796  2.142760  5.1326458 1.058629e-11 9.802907e-09
##   7:  eyes the 122 1528  62 1738  -1.1620034 -1.638598 -0.6973805 2.839285e-07 8.763925e-05
##   8:  hand his 175 2315 114 2586  -0.7778652 -1.140960 -0.4192322 1.183505e-05 4.540716e-03
##   9:  hand   i  19 2471  75 2625   1.8932508  1.146902  2.7069140 2.704759e-08 1.556589e-05
##  10:  hand  my  12 2478  86 2614   2.7637906  1.880930  3.7755453 5.884107e-14 6.772608e-11
##  11: hands  my   5 1125  45 1775   2.5113109  1.177037  4.2063750 1.127123e-05 9.321308e-03
##  12:  head  my  10 1760  62 2258   2.2723508  1.292764  3.4061853 1.024855e-07 1.111968e-04

plot(results)

Further examples of how the method has been used can be found in:

• Mahlberg, M., Wiegand, V., & Hennessey, A. (2020). Eye language – body part collocations and textual contexts in the nineteenth-century novel. In L. Fesenmeier & I. Novakova (Eds.), Phraseology and Stylistics of Literary Language/Phraséologie et Stylistique de la Langue Littéraire (pp. 143–176). Peter Lang. https://www.academia.edu/45152494/Eye_language_body_part_collocations_and_textual_contexts_in_the_nineteenth_century_novel

• Wiegand, V. (2019). A Corpus Linguistic Approach to Meaning-Making Patterns in Surveillance Discourse [PhD, University of Birmingham]. https://etheses.bham.ac.uk/id/eprint/9778

Installing from CRAN

In an R session type

install.packages('CorporaCoCo')

Installing the latest development version directly from GitHub

Linux

In an R session type:

pkg_file <- tempfile()
download.file(url = 'https://github.com/mahlberg-lab/CorporaCoCo/archive/master.tar.gz', mode = 'wb', method = 'wget', destfile = pkg_file)
install.packages(pkg_file, repos = NULL, type = 'source')

Mac OSX / Windows

download.file may not support fetching https URLs. Alternatively, you can use the the CRAN package downloader to fetch the archive instead:

# install.packages("downloader")
pkg_file <- tempfile()
downloader::download(url = 'https://github.com/mahlberg-lab/CorporaCoCo/archive/master.tar.gz', mode = 'wb', destfile = pkg_file)
install.packages(pkg_file, repos = NULL, type = 'source')

Alternatively use the devtools CRAN package

If you have the CRAN package devtools you can use this to install directly from github:

# install.packages("devtools")
devtools::install_github("mahlberg-lab/CorporaCoCo")

Testing

Unit tests are located in the /tests/testthat directory. We use the 'testthat' package to generate tests.

To run the tests yourself, just do:

devtools::test()
ℹ Loading CorporaCoCo
ℹ Testing CorporaCoCo
✔ |  OK F W S | Context
✔ |  39       | coco [0.5 s]
✔ |   4       | corp_concordance [0.1 s]
✔ |  10       | corp_cooccurrence
✔ |  12       | corp_text
✔ |   2       | surface_coco [0.1 s]
✔ |  44       | surface [0.3 s]

══ Results ══════════════════════════════════

Duration: 1.2 s

[ FAIL 0 | WARN 0 | SKIP 0 | PASS 111 ]

Continuous integration testing is set up using GitHub Actions - see the .github directory in the root of this project for more information.