If lexsync contributes to published work, please cite it. The reference below gives the DOI that CRAN assigned to the package.
Bernabeu, P. (2026). lexsync: Lexical optimisation and hardware-timed experiment generation. R package version 0.2.0. Comprehensive R Archive Network (CRAN). https://doi.org/10.32614/CRAN.package.lexsync
@Manual{lexsync,
title = {{lexsync}: Lexical optimisation and hardware-timed experiment generation},
author = {Pablo Bernabeu},
year = {2026},
publisher = {Comprehensive R Archive Network (CRAN)},
note = {R package version 0.2.0},
doi = {10.32614/CRAN.package.lexsync},
url = {https://CRAN.R-project.org/package=lexsync},
}
R users can also retrieve the reference with
citation("lexsync"), which formats it as the APA reference
for software. The 0.1.0 release on CRAN was built before the DOI existed
and prints the GitHub address, while 0.1.1 and later print the DOI.
The repository also holds a machine-readable CITATION.cff,
which GitHub turns into a ready-made citation through its ‘Cite this
repository’ button and which reference managers can import. Because it
describes both packages, it gives the concept DOI of the Zenodo archive
that holds every GitHub release of both. It also lists each package’s
identifiers, the CRAN DOI above among them.
A manuscript describing lexsync is in preparation, under the title
lexsync: A cross-platform pipeline for multidimensional lexical
optimisation and hardware-timed experiment generation. It has no
venue as yet. Once it is accepted, CITATION.cff will name
it as the preferred citation, and until then the software reference
above is the one to use.
The software citation covers the tool, not the data your items came
from. The corpora are third-party work with terms of their own, and
every one is credited, with its licence and its retrieval date, in corpora/ATTRIBUTION.md.
Each run’s materials datasheet records the source file the run read and
its SHA-256. The corpus behind a published stimulus set therefore stays
identifiable long after the run, and a reader can check that the file
they hold is the one the selection was made from.
Pablo Bernabeu is a researcher in the Department of Education at the University of Oxford, with hands-on experience of behavioural experiments, EEG, corpus analysis, computational modelling and statistics. He develops open and reproducible research software in R and Python, and is a Fellow of the Software Sustainability Institute. His ORCID record lists his other work.
lexsync has a feature-parity twin in Python, documented at its own site. The two packages are built from one repository and released in step under one version. Under the deterministic matching methods, they select byte-identical stimuli from the same lexicon and design. A group can therefore work in whichever language suits it without the materials diverging.
The source code is released under the MIT
licence. The corpus derivatives bundled with the package are not
covered by it. They inherit share-alike terms from the corpora they were
built from, so they are distributed under CC BY-SA 4.0. LICENSE-DATA
sets out the reasons, and corpora/ATTRIBUTION.md
records the sources and retrieval dates. If you have only the installed
package, its COPYRIGHTS file gives the same terms and the
credit they require, and
system.file("COPYRIGHTS", package = "lexsync") locates
it.
The distinction matters if you redistribute a lexicon derived from the bundled data, since that means crediting the original corpus authors, saying what was changed and passing the same licence on. Nothing of the sort applies to the code you write with the package.
lexsync follows semantic versioning, and the R and Python packages share one version number so that a citation identifies the same state of both. Releases are tagged on GitHub, and the changelog records what changed in each of them.
The R package is on CRAN, which keeps the source of every version it publishes. A single DOI, https://doi.org/10.32614/CRAN.package.lexsync, covers all of those versions, so the reference above names the version as well. Zenodo also archives every GitHub release, which carries both packages, under the concept DOI https://doi.org/10.5281/zenodo.22906962. That is the identifier the Python package cites, since the CRAN DOI names the R package alone.
Bugs and feature requests are welcome on the issue tracker, which is also where questions about a design or a generated experiment are best raised. The contributing guide describes the development setup for both engines and the parity rule that governs every change, and everyone taking part is asked to honour the Code of Conduct.
A report that someone can act on includes the design YAML that was
run and the run log that run_pipeline() wrote beside the
results. Between them they record the inputs and every step that ran,
which is usually enough to reproduce a problem without a further round
of questions. Where a corpus is involved, the materials datasheet from
the same run identifies it by path and checksum.