Merge branch 'master' of github.com:RasaHQ/rasa_core

docs/domains.rst

@@ -49,7 +49,7 @@ Defining Custom Actions
 
 The easiest are ``UtterActions``, which just send a message to the user. You define them by adding an entry to the
 action list that is named after the utterance. E.g. if there should be an action that utters the template called
-``utter_greet`` you need to add ``greet`` to the list of defined actions. In the above example yaml you can see that
+``utter_greet`` you need to add ``utter_greet`` to the list of defined actions. In the above example yaml you can see that
 all three of the defined actions are just named after utter templates and hence just respond with a message to
 the user.
 
@@ -66,7 +66,7 @@ your bot would execute the action ``ActionCheckRestaurants``, which might look l
    class ActionCheckRestaurants(Action):
       def name(self):
          # type: () -> Text
-         return "check_restaurants"
+         return "action_check_restaurants"
 
       def run(self, dispatcher, tracker, domain):
          # type: (Dispatcher, DialogueStateTracker, Domain) -> List[Event]
@@ -124,7 +124,7 @@ Templates defined in a domains yaml file can contain images and buttons as well:
 .. code-block:: yaml
 
    templates:
-     utter_greeting:
+     utter_greet:
      - text: "Hey! How are you?"
        buttons:
        - title: "great"
@@ -153,7 +153,7 @@ like this:
 .. code-block:: yaml
 
   templates:
-    utter_greeting:
+    utter_greet:
     - text: "Hey, {name}. How are you?"
 
 Rasa will automatically fill that variable with a value found in a slot called

docs/scheduling.rst

@@ -50,7 +50,7 @@ Here is an example implementation for our ``action_confirm_booking``:
 
     class ActionConfirmBooking(Action):
         def name(self):
-            return "confirm_booking"
+            return "action_confirm_booking"
 
         def run(self, dispatcher, tracker, domain):
             dispatcher.utter_message("Do you want to confirm your booking at Papi's pizza?")
@@ -71,7 +71,7 @@ example could look like this:
 
     class ActionBookingReminder(Action):
         def name(self):
-            return "booking_reminder"
+            return "action_booking_reminder"
 
         def run(self, dispatcher, tracker, domain):
             dispatcher.utter_message("You have an unconfirmed booking at Papi's pizza, would you like to confirm it?")

docs/tutorial_basics.rst

@@ -78,24 +78,25 @@ So what do the different parts mean?
 | ``templates`` | template strings for the things your bot can say                                                     |
 +---------------+------------------------------------------------------------------------------------------------------+
 
-In our simple example we don't need slots, so that section doesn't appear
+In our simple example we don't need ``slots`` and ``entities``, so these sections don't appear
 in our definition.
 
 **How does this fit together?**
 Rasa takes the ``intent``, ``entities``, and the internal state of the dialogue,
 and selects one of the ``actions`` that should be executed next.
 If the action is just to say something to the user, Rasa will look for a matching
 template in the domain (action name equals the utter template, as for
-``utter_greeting`` in the above example), fill in any variables,
-and respond.
+``utter_greet`` in the above example), fill in any variables,
+and respond. For actions
+which do more than just send a message, you can define them as
+python classes and reference them in the domain by their module path. See
+:ref:`custom_actions` for more information about custom actions.
 
 .. note::
 
-   There is one special action, ``ActionListen``, which means to stop taking
-   further actions until the user says something else. For actions
-   which do more than just send a message, you can define them as
-   python classes and reference them in the domain by their module path. See
-   :ref:`custom_actions` for more information about custom actions.
+   There is one additional special action, ``ActionListen``, which means to stop taking
+   further actions until the user says something else.
+   It is not specified in the ``domain.yml``
 
 2. Define an interpreter
 ^^^^^^^^^^^^^^^^^^^^^^^^
@@ -106,6 +107,7 @@ In this example we are going to use Rasa NLU for this purpose.
 
 In Rasa NLU, we need to define the user messages our bot should be able to
 handle in the `Rasa NLU training data format <https://nlu.rasa.ai/dataformat.html>`_.
+In this tutorial we are going to use Markdown Format for NLU training data.
 Let's create some intent examples in ``data/nlu.md``:
 
 .. literalinclude:: ../examples/moodbot/data/nlu.md
@@ -130,8 +132,9 @@ Let's run
 
    python -m rasa_nlu.train -c nlu_model_config.json --fixed_model_name current
 
-to train our NLU model. A new directory ``models/nlu/current`` should have been
-created containing the NLU model.
+to train our NLU model. A new directory ``models/nlu/default/current`` should have been
+created containing the NLU model. Note that ``default`` stands for project name, since we did not
+specify it explicitly in ``nlu_model_config.json``.
 
 .. note::
 
@@ -144,26 +147,32 @@ created containing the NLU model.
 So far, we've got an NLU model, a domain defining the actions our bot can
 take, and inputs it should handle (intents & entities). We are still
 missing the central piece, **stories to tell our bot what to do at which
-point in the dialogue**. There are two different ways to create stories (and
+point in the dialogue**.
+
+A **story** is a training data sample for the dialogue system.
+There are two different ways to create stories (and
 you can mix them):
 
 - create the stories by hand, writing them directly to a file
 - create stories using interactive learning (see :ref:`tutorial_interactive_learning`).
 
 For this example, we are going to create the stories by writing them directly
-to ``stories.md``. But be aware, although it is a bit faster to write
-stories directly by hand instead of using interactive learning, special
-care needs to be taken when using slots, as they need to be properly set in the
-stories. But enough talking, let's head over to our stories:
+to ``stories.md``.
+Stories begin with ``##`` and a string as an identifier. User actions start
+with an asterisk, and bot actions are specified by lines beginning with a
+dash. The end of a story is denoted by a newline. See :ref:`stories` for
+more information about the data format.
+
+Enough talking, let's head over to our stories:
 
 .. literalinclude:: ../examples/moodbot/data/stories.md
     :linenos:
     :language: md
 
-Stories begin with ``##`` and a string as an identifier. User actions start
-with an asterisk, and bot actions are specified by lines beginning with a
-dash. The end of a story is denoted by a newline. See :ref:`stories` for
-more information about the data format.
+Be aware, although it is a bit faster to write
+stories directly by hand instead of using interactive learning, special
+care needs to be taken when using slots, as they need to be properly set in the
+stories.
 
 4. Put the pieces together
 ^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -183,13 +192,18 @@ Here we'll just talk to the bot on the command line:
 
 .. code-block:: bash
 
-   python -m rasa_core.run -d models/dialogue -u models/nlu/current
+   python -m rasa_core.run -d models/dialogue -u models/nlu/default/current
 
 And there we have it! A minimal bot containing all the important pieces of
 Rasa Core.
 
 .. image:: _static/images/facebook-run.png
 
+.. note::
+
+    Button emulation does not work in console output, you need to type words like "great" or "sad"
+    instead of numbers 1 or 2.
+
 Bonus: Handle messages from Facebook
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 

docs/tutorial_interactive_learning.rst

@@ -75,11 +75,11 @@ Below is an excerpt of the stories.
 
    ## greet
    * greet
-       - action_greet
+       - utter_greet
 
    ## happy
    * thankyou
-       - action_youarewelcome
+       - utter_youarewelcome
    ...
 
    ## compare_reviews_venues
@@ -94,8 +94,7 @@ Below is an excerpt of the stories.
 Interactive Learning
 ^^^^^^^^^^^^^^^^^^^^
 
-Run the script ``train_online.py``. This first runs the
-``train_init`` script, creating a stateless policy by combining the stories
+Run the script ``train_online.py``. This first creats a stateless policy by combining the stories
 we've provided into longer dialogues, and then trains the policy on that
 dataset.
 
@@ -110,7 +109,7 @@ where the learning *becomes interactive*):
     so when you type messages to the bot you have to
     type the intent starting with a ``/`` (see :ref:`fixed_intent_format`).
     If you want to use Rasa NLU / wit.ai / Lex you
-    can just swap the ``Interpreter`` class in ``run.py``.
+    can just swap the ``Interpreter`` class in ``run.py`` and ``train_online.py``.
 
 We now start talking to the bot by directly entering the intents. For
 example, if we type ``/greet``, we get the following prompt:
@@ -121,13 +120,15 @@ example, if we type ``/greet``, we get the following prompt:
    ------
    Chat history:
 
+        bot did:    None
         bot did:	action_listen
         user said:	/greet
-        	   whose intent is:	greet
 
-   we currently have slots: {'location': None}
+                whose intent is:	greet
+
+   we currently have slots: concerts: None, venues: None
    ------
-   The bot wants to [greet] due to the intent. Is this correct?
+   The bot wants to [utter_greet] due to the intent. Is this correct?
 
        1.	Yes
        2.	No, intent is right but the action is wrong
@@ -137,13 +138,14 @@ example, if we type ``/greet``, we get the following prompt:
 
 This gives you all the info you should hopefully need to decide
 what the bot *should* have done. In this case, the bot chose the right
-action ('greet'), so we type ``1`` and hit enter. We continue this loop
-until the bot chooses the wrong action.
+action ('utter_greet'), so we type ``1`` and hit enter.
+Then we type ``1`` again, because 'action_listen' is the correct action after greeting.
+We continue this loop until the bot chooses the wrong action.
 
 **Providing feedback on errors**
 
-We've just asked the bot to search for concerts, and now we're asking it to
-compare reviews. The bot happens to choose the wrong one out of the two
+If you ask ``/search_concerts``, the bot should suggest ``action_search_concerts`` and then ``action_listen``.
+Now let's ask it to ``/compare_reviews``. The bot happens to choose the wrong one out of the two
 possibilities we wrote in the stories:
 
 .. code-block:: text
@@ -153,15 +155,14 @@ possibilities we wrote in the stories:
    Chat history:
 
         bot did:	action_search_concerts
-        bot did:	action_suggest
         bot did:	action_listen
         user said:	/compare_reviews
 
         	   whose intent is:	compare_reviews
 
-   we currently have slots: {'location': None}
+   we currently have slots: concerts: [{'artist': 'Foo Fighters', 'reviews': 4.5}, {'artist': 'Katy Perry', 'reviews': 5.0}], venues: None
    ------
-   The bot wants to [show_venue_reviews] due to the intent. Is this correct?
+   The bot wants to [action_show_venue_reviews] due to the intent. Is this correct?
 
        1.	Yes
        2.	No, intent is right but the action is wrong
@@ -176,22 +177,21 @@ model has assigned to each of the actions.
 .. code-block:: text
    what is the next action for the bot?
 
-        0	default	 0.00148131744936
-        1	greet	 0.0970264300704
-        2	goodbye	 0.0288009047508
-        3	listen	 0.00123148341663
-        6	search_cinemas	0.000627864559647
-        8	search_films	0.0367559418082
-        9	suggest		0.0261212754995
-        11	youarewelcome	0.594935178757
-        13	explain_options	0.0516758263111
-        14	store_slot	0.00145904591773
-        15	show_cinema_reviews	0.00887114647776
-        16	show_film_reviews	0.0870243906975
+        0                           action_listen    0.19
+        1                          action_restart    0.00
+        2                           utter_default    0.00
+        3                             utter_greet    0.03
+        4                           utter_goodbye    0.03
+        5                     utter_youarewelcome    0.02
+        6                  action_search_concerts    0.09
+        7                    action_search_venues    0.02
+        8             action_show_concert_reviews    0.29
+        9               action_show_venue_reviews    0.33
+
 
 
-In this case, the bot should ``show_film_reviews`` (rather than cinema
-reviews!) so we type ``16`` and hit enter.
+In this case, the bot should ``action_show_concert_reviews`` (rather than venue
+reviews!) so we type ``8`` and hit enter.
 
 .. note::
 
@@ -206,6 +206,15 @@ current conversation to a file and exit the conversation. Make sure to
 combine the dumped story with your original training data for the next
 training.
 
+.. note::
+
+    If you run the bot with not enough training data, it might get ``action_listen``
+    as a most probable response to your input and therefore do nothing.
+    If you continue to input something and get no answer, please head to
+    interactive training and check if ``action_listen`` was chosen as a response.
+    Correct the bot's behaviour, add additional stories and run ``train.py`` then
+    run the bot again.
+
 Motivation: Why Interactive Learning?
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^