Maintenance maintenance

Ideally, each maintenance release requires no more than 50% the time and effort of the previous one, which ensures reasonable convergence.

v0.2.5.1

Maintenance update, nothing to see here.

v0.2.5

Some support for visualization of multiple models together (sort of) using dfr-browser v0.8.1-alpha.

v0.2.4.3

A tiny incremental release with no new features, to remove the python dependency. Python was used only for converting the Gibbs sampling state output by MALLET into a format we could use; this is now done in R (a little more slowly). load_from_mallet_state should function on a slightly wider range of configurations now.

v0.2.4.2

A release with further small bug fixes and tweaks. In particular a problem with load_from_mallet_state and the most recent bigtabulate package is worked around here. I've also added a permute parameter for reordering topic numbers in a dfr_browser export (see ?dfr_browser and export_browser_data for some detail).

v0.2.4.1

Miscellaneous minor added features (some added since 0.2.4, which I never made into a release here):

• foreign_model, for using basic package functions on topic models from topicmodels and stm
• a proper parameter to dfr_browser and export_browser_data, which, if set, indicates that the document-topic and topic-word matrices should be smoothed and normalized
• a dfrtopics.mallet_logging option to control MALLET's verbose logging output, since the usual methods (capture.output, suppressMessages) do not work: see help("mallet-logging")
• the included dfr-browser files have been updated
• various minor bug fixes and documentation updates.

The model alignment and posterior predictive checking functions remain in an experimental state. They produce plausible-seeming outputs, but they should be systematically double-checked.

v0.2.3

• An adjusted dfr-browser export via dfr_browser() for one-line interactive browsing. A slightly updated dfr-browser is also included with the package. export_browser_data is still available.
• wordcounts_instances to wrap the wordcounts_texts() %>% make_instances() usage with the right defaults for no further tokenizing/stopping/casefolding.
• Various vignette updates.