NEWS

---

LSX 1.4.1

• Add group to smooth_lss() to smooth LSS scores by group.
• Add optimize_lss() as an experimental function.
• Add prop_slice to textmodel_lss() an experimental argument.
• Deprecate auto_weight in textmodel_lss().

LSX 1.4.0 (2024-03-05)

• Change the default value to max_highlighted = 1000 in textplot_terms().
• Add ... to customize text labels to textplot_terms().
• Highlight words in different colors when a dictionary is passed to highlighted.
• Add mode = "predict" and remove = FALSE to bootstrap_lss().

LSX 1.3.2 (2023-12-20)

• Fix the error in textplot_terms() when the frequency of terms are zero (#85).

LSX 1.3.1 (2023-02-26)

• Fix the range of scores when cut is used.
• Add bootstrap_lss() as an experimental function.

LSX 1.3.0 (2023-01-22)

• Add cut to predict.
• Move examples to the new package website: http://koheiw.github.io/LSX.
• Rename "rescaling" to "rescale" for simplicity and consistency.
• Improve random sampling of words to highlight in textplot_terms() to avoid congestion.

LSX 1.2.0 (2022-12-04)

• Add group_data to textmodel_lss() to simplify the workflow.
• Add max_highlighted to textplot_terms() to automatically highlight polarity words.

LSX 1.1.4

• Update as.textmodel_lss() to avoid errors in textplot_terms() when terms is used.

LSX 1.1.3 (2022-10-19)

• Restore examples for textmodel_lss().
• Defunct char_keyness() that has been deprecated for long.

LSX 1.1.2 (2022-10-02)

• Update examples to pass CRAN tests.

LSX 1.1.1 (2022-02-26)

• Add min_n to predict() to make polarity scores of short documents more stable.

LSX 1.1.0 (2022-02-24)

• Add as.textmodel_lss() for textmodel_lss objects to allow modifying existing models.
• Allow terms in textmodel_lss() to be a named numeric vector to give arbitrary weights.

LSX 1.0.2 (2021-09-18)

• Add the auto_weight argument to textmodel_lss() and as.textmodel_lss() to improve the accuracy of scaling.
• Remove the group argument from textplot_simil() to simplify the object.
• Make as.seedwords() to accept multiple indices for upper and lower.

LSX 1.0.0 (2021-07-20)

• Add max_count to textmodel_lss.fcm() that will be passed to x_max in rsparse::GloVe$new().
• Add max_words to textplot_terms() to avoid overcrowding.
• Make textplot_terms() to work with objects from textmodel_lss.fcm().
• Add concatenator to as.seedwords().

LSX 0.9.9 (2021-04-19)

• Correct how textstat_context() and char_context() computes statistics.
• Deprecate char_keyness().

LSX 0.9.8 (2021-03-22)

• Stop using functions and arguments deprecated in quanteda v3.0.0.

LSX 0.9.7 (2021-03-08)

• Make as.textmodel_lss.matrix() more reliable.
• Remove quanteda.textplots from dependencies.

LSX 0.9.6 (2020-12-17)

• Updated to reflect changes in quanteda (creation of quanteda.textstats).

LSX 0.9.4 (2020-11-02)

• Fix char_context() to always return more frequent words in context.
• Experimental textplot_factor() has been removed.
• as.textmodel_lss() takes a pre-trained word-embedding.

LSX 0.9.3

• Add textstat_context() and char_context() to replace char_keyness().
• Make the absolute sum of seed weight equal to 1.0 in both upper and lower ends.
• textplot_terms() takes glob patterns in character vector or a dictionary object.
• char_keyness() no longer raise error when no patter is found in tokens object.
• Add engine to smooth_lss() to apply locfit() to large datasets.

LSX 0.9.2 (2020-09-22)

• Updated unit tests for the new versions of stringi and quanteda.

LSX 0.9.0 (2020-09-09)

• Renamed from LSS to LSX for CRAN submission.

LSX 0.8.7

• Added textplot_terms() to improve visualization of model terms.