GitLab

Latino: Added widgets dealing with vocabulary from CrossBee. Widget BOW Space changed to BOW model and doesn't generate dataset anymore. Now it needs to be followed by Create Dataset. Fine tuning widgets interfaces based on PhD workflow and widget names suggested by Nada. Few new widgets still don't have correct background logic.

CrossBee: Moved widgets dealing with vocabulary to Latino. Fine tuning widgets interfaces based on PhD workflow and widget names suggested by Nada. Many widgets still don't have background logic.

Latino: Some widgets' logic changed (e.g. Bow Model now generates dataset and dataset cannot be generated in one step anymore)

nicer tooltip display for widget inputs and outputs; I don't know how to make it refresh/apply after adding a new widget

Latino: 2 new widgets - for joining document corpora and for raw document preprocessing (aka raw bow generation) - implementation not final yet.

CrossBee: 9 new widgets dealing mainly with heuristic calculation, results exploration and methodlolgy evaluation. Reorganization of categories.

LemmaGen: different python.net loading style - switching to "late binding" which means python.net needs to be installed inside python and 'import clr' need to be called in .py files.

CrossBee: different python.net loading style - switching to "late binding" which means python.net needs to be installed inside python and 'import clr' need to be called in .py files.

CrossBee: main feature is new outlier detection widgets along with outlier heuristics, debuged other functionality and added categories and reorganized widgets

LemmaGen: Four new widgets added for saving/loading objects to files, reorganized into new categories & connected to new LATINO dlls build: based on latino ver. 9d91c3e (main change is empty vector support in cosine similarity)

* New LATINO dlls build: based on latino ver. 9d91c3e (main change is empty vector support in cosine similarity)
* New widgets: Dataset to python object, Python object to dataset, Advanced Object to String Converter
* Corected many bugs related to empty vector used in classification

Housekeeping: some action functions ('base' package - Objects category) moved from library.py to base\library.py.

wordification "sentence" caching
dbcontext bug ko sta dve entiti imeli lahko le eno relacijo med njima

CrossBee: First 9 widgets added, tested and debugged - mainly heuristic definintion/creation functionality

Latino: considerably improved accuracy calculation which is due to complexity now implemented in c# and not in python as before.