2.0

Requirement: Java 8

New Features:

• All-in-one command for hierarchical topic detection
• Webpage visualization with direct link to corresponding documents
• Evaluation metrics: topic coherence, topic compactness(scala ver.)
• Allow input document to be listed line by line
• Supports non-ascii characters
• Supports LDA data format
• Added option to skip tree level
• Simplified HLTA parameters
• Supports seedwords of any word length
• Parallel computation in computing word-pair MI

Other changes:

• Default using Narrowly Defined Topics
• Scala calls use Stepwise EM for parameter estimation
• User defined encoding scheme in data conversion
• Pre-processor now remove punctuation instead of replacing it with underscore
• Subroutines now accept all data formats, while sparse data will be the default format
• Data Conversion default only outputs sparse data format
• Data Conversion now reads PDF directly
• Sparse data format now counts docId from 0
• HLCM data format now uses extension .hlcm
• Legacy fixes of collision with .bif format reserved words
• Fixed invalid json format