Changes in v0.6.0

• Adds predict() to identify topics of unseen documents (#9)
• Allows selecting seed words based on their frequencies using dfm_trim() in textmodel_seededlda() via ... (#8)

Changes in v0.5.1

• topics() now returns factor with NA for empty documents
• Fix a bug in initializing LDA that leads to incorrect phi (#4 and #6)

Changes in v0.5

• Implement original LDA estimator using the LDAGibbs++ library