NEWS.md

In juba/rainette: The Reinert Method for Textual Data Clustering

rainette (development version)

• Add colors argument to rainette_plot() to allow to specify custom colors for clusters (#37, thanks @lvaudor)

rainette 0.3.1

• Replace call to palette.colors() with manual colors vector to ensure compatibility with R 3.6 (#11)
• Fix dendrogram not displayed when show_na_title = FALSE

rainette 0.3.0

New features

Rework of the double classification computation :

• New full argument to rainette2(). If TRUE (default, same behavior as previous versions), all crossings between groups of the two single classifications are taken into account to find the best partition. If FALSE, only the crossings with maximal associations are kept.
• New parallel argument to rainette2() to compute partitions with mclapply (FALSE by default, won't work on Windows, uses more RAM)
• Global optimization and speed up of rainette2() computations, with added progress bars to better estimate long runs

Documentation

• Improved french vignette "description des algorithmes"
• New english vignette "algorithms description"
• Reworked french and english introduction vignettes

Deprecated features

• The wordcloud plots will be deprecated in a near future. A warning has been added to rainette_plot() and rainette2_plot() if they are called with type = "cloud".

Other

• Add show_na_title, cluster_label and keyness_plot_xlab arguments to rainette_plot() to customize graphics output
• Fix warnings in Font Awesome icon names

rainette 0.2.1

• Add option to show merged segments in document browser
• Fix warning and error in rainette_explor and rainette2_explor when a cluster dfm is empty
• Fix error when the dfm contains empty string as feature

rainette 0.2.0

Important and breaking changes

• min_uc_size, uc_size1 and uc_size2 arguments to rainette and rainette2 have been renamed to min_segment_size, min_segment_size1 and min_segment_size2.
• The default value of min_segment_size in rainette is now 0, which means that no merging is done between segments by default. Results could then be different from previous package versions when min_uc_size was not specified.
• Important bugfix : merging of segments based on min_segment_size was not handled correctly in the previous versions regarding the segment sources, as segments from different documents could be merged together. This should now be fixed.

New features

• A new graphical interface to browse cluster documents has been added to rainette_explor and rainette2_explor.
• New function clusters_by_doc_table which gives the number of segments of each cluster for each document.
• New function docs_by_cluster_table which gives, for each cluster, the number of documents with at least one segment in this cluster.
• split_segments should now be about 4 times faster.
• Terms frequencies and documents proportions statistics have been added to the explor interfaces.

Other

• When rainette is called with min_segment_size > 0, a doc_id argument must be given which is the name of a dtm docvar identifying the segments source. If the corpus has been produced by split_segments, the added segment_source docvar is used by default.
• Color palette for clusters changed to "Tableau 10".
• Negative keyness values are not shown by default anymore in rainette_explor, rainette2_explor, rainette_plot and rainette2_plot.
• Wordcloud plots have been removed from explor interfaces.
• A warning is displayed when min_split_members < 3.
• If rainette_explor is called on a rainette2 results object, rainette2_explor is launched automatically.

rainette 0.1.3

• Parallel computing in split_segments has been removed
• Fix potential name conflicts in rainette_explor
• Compatibility with quanteda v3 (thanks @kbenoit)

rainette 0.1.2

• Fix bug due to factor comparison in rainette2

rainette 0.1.1

• Compatibility with dplyr 1.0
• Fix bug in p threshold in rainette_stats

rainette 0.1.0

• First version

juba/rainette documentation built on May 9, 2024, 4:27 p.m.