- Add
boolean
totextmodel_newsmap()
.
- Add Turkish seed dictionary
- Add
as.dictionary()
fortextmodel_newsmap
.
- Add
select
tocoef()
and improve its documentation
- Fix for changes in the Matrix package v1.5
- Add
min_conf
topredict()
- Add experimental argument
entropy
totextmodel_newsmap()
- Update the seed words for UA (#64)
- Add Portuguese dictionary
- Add "Saigon" to VE in selected languages (#47)
- Add
label
anddrop_label
totextmodel_newsmap()
- Add
rescale
andmin_n
topredict()
- Add
print()
methods for fitted models
- Change predicted values to factor with all labels in training data
- Update tests for quanteda v3.0
- Make
predict()
significantly faster
- Improve efficiency of
textmodel_newsmap()
- Add compatibility with newer
textstat_entropy()
- Add Hebrew and Arabic seed dictionaries
- Clean up Italian, German and Spanish seed dictionaries
- Add Italian seed dictionary
- Correct Japanese seed words for DE, MG and EC
- Return NA for documents that do not have known features
- Drop document variables to avoid slowdown and warnings
- Add Chinese (simplified and traditional) seed dictionaries
- Add French seed dictionary
- Add a function to compute average feature entropy (AFE)
- Fix error in
textmodel_newsmap()
when smooth is < 1.0 - Fitted models no longer include classes that did not occur in training set
- Add
coef()
andcoefficients()
methods
- Add Russian language seed dictionary
- Correct Cook Islands' country code from CC to CK, and remove it from POL
- De facto capital cities of TZ, ZA, NG and BO are added to all dictionaries
- CC is added to DE, JA and EN dictionaries
- Dictionary entries are sorted in alphabetical order of country code