Merge 1318e67 into 3a3b290

.travis.yml

@@ -40,7 +40,7 @@ jobs:
     install:
     - pip install sphinx==1.7.5 sphinx-autobuild==0.7.1 sphinxcontrib-programoutput==0.11 nbsphinx==0.3.2 sphinxcontrib-httpdomain==1.5.0
     - pip install git+https://${GITHUB_TOKEN}:x-oauth-basic@github.com/RasaHQ/sphinxcontrib-versioning.git@version_list
-    - pip install git+https://${GITHUB_TOKEN}:x-oauth-basic@github.com/RasaHQ/rasabaster.git#egg=rasabaster
+    - pip install https://storage.googleapis.com/docs-theme/rasabaster-0.7.10.tar.gz
     - pip install -e git://github.com/RasaHQ/sphinx_rtd_theme.git#egg=sphinx_rtd_theme
     - pip install -e .
     - pip list

alt_requirements/requirements_dev.txt

@@ -20,4 +20,5 @@ sphinx-autobuild==0.7.1
 sphinxcontrib-versioning==2.2.1
 sphinxcontrib-programoutput==0.11
 nbsphinx==0.2.18
+https://storage.googleapis.com/docs-theme/rasabaster-0.7.10.tar.gz
 -e git://github.com/RasaHQ/sphinx_rtd_theme.git#egg=sphinx_rtd_theme

docs/entities.rst

@@ -60,7 +60,7 @@ Regular Expressions (regex)
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
 You can use regular expressions to help the CRF model learn to recognize entities.
-In the :ref:`dataformat` you can provide a list of regular expressions, each of which provides
+In the :ref:`section_dataformat` you can provide a list of regular expressions, each of which provides
 the ``ner_crf`` with an extra binary feature, which says if the regex was found (1) or not (0). 
 
 For example, the names of German streets often end in ``strasse``. By adding this as a regex,

docs/fallback.rst

@@ -3,3 +3,30 @@
 Confidence and Fallback Intents
 ===============================
 
+
+Each of the pipelines will report a ``confidence`` score along with the predicted intent,
+and the ``ner_crf`` component will do the same for the extracted entities. 
+
+You can use the confidence score to choose when to ignore Rasa NLU's prediction and trigger
+fallback behaviour, for example asking the user to rephrase. If you are using Rasa Core,
+you can do this using a `Fallback Policy </core/fallbacks/>`_. 
+
+Choosing a Confidence Cutoff
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+A good way to choose a confidence cutoff is to calculate the model's confidence on a test set,
+and compare the confidence values on the correctly and incorrectly predicted examples.
+
+A Note about Confidence Scores
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Always keep in mind that the confidence score is not a true probability that the prediction 
+is correct, it's just a metric defined by the model that approximately describes how similar
+your input was to the training data. 
+
+The intent classifier in the ``spacy_sklearn`` pipeline, for example, usually reports very low
+confidence numbers, whereas the ``tensorflow_embedding`` pipeline usually provides very high confidences.
+One common misconception is that if your model reports high confidence on your training examples,
+it is a "better" model. In fact, this usually means that your model is overfitting.
+
+

docs/pipeline.rst

@@ -6,9 +6,10 @@ Pipeline and Component Configuration
 .. contents::
 
 
-The process of incoming messages is split into different components. These components are executed one after another
+Incoming messages are processed by a sequence of components. These components are executed one after another
 in a so called processing pipeline. There are components for entity extraction, for intent classification,
-pre-processing and there will be many more in the future.
+pre-processing, and others. If you want to add your own component, for example to run a spell-check or to
+do sentiment analysis, check out :ref:`section_customcomponents`. 
 
 Each component processes the input and creates an output. The ouput can be used by any component that comes after
 this component in the pipeline. There are components which only produce information that is used by other components