update models

Makefile

@@ -28,7 +28,7 @@ train-core:
 	python -m rasa_core.train -d domain.yml -s data/stories.md -o models/current/dialogue -c core_config.yml
 
 interactive:
-	python -m rasa_core.train interactive --core models/current/dialogue -d domain.yml -c core_config.yml
+	python -m rasa_core.train interactive --core models/current/dialogue -d domain.yml -c core_config.yml -u models/current/nlu --endpoints endpoints.yml
 
 cmdline:
 	python -m rasa_core.run -d models/current/dialogue -u models/current/nlu --endpoints endpoints.yml

data/nlu_data.md

@@ -1,5 +1,13 @@
-## intent:bye <!--- The label of the intent --> 
-- Bye 			<!--- Training examples for intent 'bye'--> 
+## intent:affirm
+- yes
+- yes sure
+- absolutely
+- for sure
+- yes yes yes
+- definitely
+
+## intent:bye
+- Bye
 - Goodbye
 - See you later
 - Bye bot
@@ -31,6 +39,16 @@
 - hello everybody
 - hello is anybody there
 - hello robot
+- hi
+
+## intent:inform
+- [10119](number)
+
+## intent:search_hospital
+- find me a nearby hospital
+- show me hospitals
+- i am looking for a clinic
+- hospital
 
 ## intent:thank
 - Thanks
@@ -48,16 +66,3 @@
 - amazing, thanks
 - cool, thanks
 - cool thank you
-
-## intent:affirm
-- yes
-- yes sure
-- absolutely
-- for sure
-- yes yes yes
-- definitely
-
-## intent:search_hospital
-- find me a nearby hospital
-- show me hospitals
-- i am looking for a clinic

data/nlu_interactive.md

@@ -0,0 +1,5 @@
+## intent:greet
+- hi
+
+## intent:search_hospital
+- hospital

data/stories.md

@@ -11,7 +11,7 @@
  - utter_greet
 
 ## happy path
-* request_hospital
+* search_hospital
     - hospital_form
     - form{"name": "hospital_form"}
     - form{"name": null}
@@ -21,9 +21,21 @@
 ## happy path 2
 * greet
     - utter_greet
-* request_hospital
+* search_hospital
     - hospital_form
     - form{"name": "hospital_form"}
     - form{"name": null}
 * thankyou
     - utter_noworries
+## Generated Story -6770035372518111269
+* greet
+    - utter_greet
+* search_hospital
+    - hospital_form
+    - form{"name": "hospital_form"}
+    - slot{"requested_slot": "zip"}
+* form: inform{"number": "10119"}
+    - form: hospital_form
+    - slot{"zip": "10119"}
+    - slot{"requested_slot": "specialty"}
+

domain.yml

@@ -1,11 +1,14 @@
 intents:
 - greet
 - goodbye
-- thanks
+- thankyou
 - deny
+- search_hospital
+- inform
 
 entities:
 - specialty
+- number
 
 slots:
   zip:

models/current/dialogue/domain.json

@@ -3,7 +3,8 @@
     "intent_deny",
     "intent_goodbye",
     "intent_greet",
-    "intent_thanks",
+    "intent_search_hospital",
+    "intent_thankyou",
     "entity_specialty",
     "prev_action_listen",
     "prev_action_restart",

models/current/dialogue/domain.yml

@@ -13,11 +13,13 @@ entities:
 forms:
 - hospital_form
 intents:
-- deny:
-    use_entities: true
 - greet:
     use_entities: true
-- thanks:
+- search_hospital:
+    use_entities: true
+- thankyou:
+    use_entities: true
+- deny:
     use_entities: true
 - goodbye:
     use_entities: true

models/current/dialogue/policy_0_MemoizationPolicy/featurizer.json

@@ -1 +1 @@
-{"py/object": "rasa_core.featurizers.MaxHistoryTrackerFeaturizer", "use_intent_probabilities": false, "state_featurizer": {"py/object": "rasa_core.featurizers.SingleStateFeaturizer", "slot_feature_len": null, "user_feature_len": null}, "remove_duplicates": true, "max_history": 5}
+{"state_featurizer": {"user_feature_len": null, "slot_feature_len": null, "py/object": "rasa_core.featurizers.SingleStateFeaturizer"}, "use_intent_probabilities": false, "max_history": 5, "py/object": "rasa_core.featurizers.MaxHistoryTrackerFeaturizer", "remove_duplicates": true}

models/current/dialogue/policy_1_FormPolicy/featurizer.json

@@ -1 +1 @@
-{"py/object": "rasa_core.featurizers.MaxHistoryTrackerFeaturizer", "use_intent_probabilities": false, "state_featurizer": {"py/object": "rasa_core.featurizers.SingleStateFeaturizer", "slot_feature_len": null, "user_feature_len": null}, "remove_duplicates": true, "max_history": 2}
+{"state_featurizer": {"user_feature_len": null, "slot_feature_len": null, "py/object": "rasa_core.featurizers.SingleStateFeaturizer"}, "use_intent_probabilities": false, "max_history": 2, "py/object": "rasa_core.featurizers.MaxHistoryTrackerFeaturizer", "remove_duplicates": true}

models/current/dialogue/policy_metadata.json

@@ -1,31 +1,31 @@
 {
-  "trained_at": "20181121-155224",
-  "rasa_core": "0.13.0a1",
+  "python": "3.5.4",
+  "policy_names": [
+    "rasa_core.policies.memoization.MemoizationPolicy",
+    "rasa_core.policies.form_policy.FormPolicy"
+  ],
   "ensemble_name": "rasa_core.policies.ensemble.SimplePolicyEnsemble",
   "action_fingerprints": {
-    "utter_greet": {
+    "hospital_form": {
       "slots": []
     },
-    "action_listen": {
+    "utter_greet": {
       "slots": []
     },
     "utter_noworries": {
       "slots": []
     },
-    "utter_goodbye": {
+    "action_listen": {
       "slots": []
     },
-    "hospital_form": {
+    "utter_goodbye": {
       "slots": []
     }
   },
-  "policy_names": [
-    "rasa_core.policies.memoization.MemoizationPolicy",
-    "rasa_core.policies.form_policy.FormPolicy"
-  ],
+  "rasa_core": "0.13.0a1",
+  "trained_at": "20181121-161139",
   "max_histories": [
     5,
     2
-  ],
-  "python": "3.5.4"
+  ]
 }

models/current/nlu/checkpoint

@@ -0,0 +1,2 @@
+model_checkpoint_path: "/Users/alan/Developer/dialog/HCA/hospital_search/models/current/nlu/intent_classifier_tensorflow_embedding.ckpt"
+all_model_checkpoint_paths: "/Users/alan/Developer/dialog/HCA/hospital_search/models/current/nlu/intent_classifier_tensorflow_embedding.ckpt"

models/current/nlu/metadata.json

@@ -1,30 +1,18 @@
 {
-    "trained_at": "20181121-154957",
-    "language": "en",
     "training_data": "training_data.json",
+    "language": "en",
+    "rasa_nlu_version": "0.13.5",
+    "trained_at": "20181121-160804",
     "pipeline": [
         {
-            "name": "nlp_spacy",
-            "model": "en",
-            "case_sensitive": false,
-            "class": "rasa_nlu.utils.spacy_utils.SpacyNLP"
-        },
-        {
-            "name": "tokenizer_spacy",
-            "class": "rasa_nlu.tokenizers.spacy_tokenizer.SpacyTokenizer"
-        },
-        {
-            "name": "intent_featurizer_spacy",
-            "class": "rasa_nlu.featurizers.spacy_featurizer.SpacyFeaturizer"
-        },
-        {
-            "regex_file": "regex_featurizer.json",
-            "name": "intent_entity_featurizer_regex",
-            "class": "rasa_nlu.featurizers.regex_featurizer.RegexFeaturizer"
+            "class": "rasa_nlu.tokenizers.whitespace_tokenizer.WhitespaceTokenizer",
+            "name": "tokenizer_whitespace"
         },
         {
+            "max_iterations": 50,
             "BILOU_flag": true,
-            "L2_c": 0.1,
+            "classifier_file": "crf_model.pkl",
+            "L1_c": 0.1,
             "features": [
                 [
                     "low",
@@ -50,35 +38,60 @@
                     "upper"
                 ]
             ],
-            "L1_c": 0.1,
-            "name": "ner_crf",
-            "classifier_file": "crf_model.pkl",
-            "max_iterations": 50,
-            "class": "rasa_nlu.extractors.crf_entity_extractor.CRFEntityExtractor"
+            "class": "rasa_nlu.extractors.crf_entity_extractor.CRFEntityExtractor",
+            "L2_c": 0.1,
+            "name": "ner_crf"
         },
         {
-            "name": "ner_synonyms",
             "class": "rasa_nlu.extractors.entity_synonyms.EntitySynonymMapper",
-            "synonyms_file": "entity_synonyms.json"
+            "synonyms_file": "entity_synonyms.json",
+            "name": "ner_synonyms"
+        },
+        {
+            "OOV_words": [],
+            "max_ngram": 1,
+            "stop_words": null,
+            "lowercase": true,
+            "min_ngram": 1,
+            "max_df": 1.0,
+            "min_df": 1,
+            "featurizer_file": "intent_featurizer_count_vectors.pkl",
+            "token_pattern": "(?u)\\b\\w\\w+\\b",
+            "strip_accents": null,
+            "max_features": null,
+            "class": "rasa_nlu.featurizers.count_vectors_featurizer.CountVectorsFeaturizer",
+            "OOV_token": null,
+            "name": "intent_featurizer_count_vectors"
         },
         {
-            "scoring_function": "f1_weighted",
-            "C": [
-                1,
-                2,
-                5,
-                10,
-                20,
-                100
+            "batch_size": [
+                64,
+                256
             ],
-            "class": "rasa_nlu.classifiers.sklearn_intent_classifier.SklearnIntentClassifier",
-            "name": "intent_classifier_sklearn",
-            "classifier_file": "intent_classifier_sklearn.pkl",
-            "kernels": [
-                "linear"
+            "epochs": 300,
+            "intent_tokenization_flag": false,
+            "C2": 0.002,
+            "embed_dim": 20,
+            "num_neg": 20,
+            "classifier_file": "intent_classifier_tensorflow_embedding.ckpt",
+            "use_max_sim_neg": true,
+            "evaluate_on_num_examples": 1000,
+            "evaluate_every_num_epochs": 10,
+            "class": "rasa_nlu.classifiers.embedding_intent_classifier.EmbeddingIntentClassifier",
+            "hidden_layer_size_a": [
+                256,
+                128
             ],
-            "max_cross_validation_folds": 5
+            "num_hidden_layers_b": 0,
+            "droprate": 0.2,
+            "mu_pos": 0.8,
+            "intent_split_symbol": "_",
+            "hidden_layer_size_b": [],
+            "C_emb": 0.8,
+            "similarity_type": "cosine",
+            "num_hidden_layers_a": 2,
+            "mu_neg": -0.4,
+            "name": "intent_classifier_tensorflow_embedding"
         }
-    ],
-    "rasa_nlu_version": "0.13.5"
+    ]
 }