• LDAResults improvements including speedups and better memory management
• Split up the streamers modules for easier dependency and imports.

new parallel_easy utils for memory friendly iterator functionality
new threading_easy utls for easy multi_threading
VW methods are parallelized for generic text streamers
protected import statements for non-standard libs

• new TextStreamer class to handle general text stream processes
• explicit doc path passing option in TextFileStreamer
• updated version of nlp.word_tokenize
• minor bug fixes

Major improvements to the modeling.eda module.

• improved LDA predict function
• improved documentation
• removed old dependencies