Released to accompany version 2.0.54 of the NLU service. Note the refactorings in the service's changelog. Below are the external API changes.
Removed the online training endpoints.
Removed the similar word entity extractor.
Added a 'test' end point to generate a model's test report without needing to re-train it.
2.0.52
==============
No API changes, but added some examples and service (now 2.0.52) updated to include additional sentence encoders and text classifier algorithms.
2.0.47
==============
Added user_id and session_id to the retrieve endpoints.
Added missing post endpoints that provide an alternative to create a model and specify the new model's name in the url.
Added a root dashboard as well as vision and audio versions.
Added missing from and to fields to duckling endpoints' swagger spec.
Added a simple text pipeline that for now just puts multiple models behind one retrieve endpoint.
2.0.46
==============
configuration.host should no longer contain the service and version postfix, for example, '/nlu/v2' in 'https://nlu.playground.feersum.io:443/nlu/v2'. The host should now just be, for example, 'https://nlu.playground.feersum.io:443'.
2.0.45
==============
Added the text and image labelled datasets endpoints to allow a user to create and manage labelled datasets.
Added a 'has_labels' flag to the model details returned on the dashboard. Useful to know which model types have a 'get_labels' endpoint.
2.0.44
==============
Added the approximate (fuzzy) string matching text classifier.
Added the softmax temperature as a hyper parameter to the text, intent and FAQ classifiers.