Merge pull request #1 from mrkmcknz/QOL/single-file-init

[WIP] Single JSON file initialization.

isac/engine.py

@@ -1,6 +1,7 @@
 import re
 import heapq
 
+from isac.training_engine import Trainer
 from isac.entity_engine import Entity
 from isac.intent_engine import IntentBuilder
 from isac.parser_engine import Parser
@@ -10,12 +11,14 @@
 
 class Engine(object):
 
-    def __init__(self, training, tokenizer=None, trie=None):
+    def __init__(self, training_data=None, tokenizer=None, trie=None):
         self.tokenizer = tokenizer or Tokenizer()
         self.trie = trie or Trie()
-        self.training = training
-        self.tagger = Entity(self.trie, self.tokenizer, self.training)
+        self.training_data = training_data
+        self.tagger = Entity(self.trie, self.tokenizer, self.training_data)
         self.intent_parsers = []
+        if training_data:
+            Trainer(self).build()
 
     def _max_intent(self, parse_result, utterance):
 
@@ -56,21 +59,6 @@ def training_nouns(self, value, entity_type):
             if entity_type not in self.nn_training:
                 self.nn_training.append(entity_type)
 
-    def pos_training(self, data):
-
-        for tag in data['tags']:
-
-            for pos in self.tokenizer.tagger(data['text']):
-                if pos[0] == tag['value']:
-                    tag['pos'] = pos[1]
-
-        return data
-
-    def register_training(self, data):
-
-        for tag in data.get('tags'):
-            self.register_entity(tag['value'], tag['key'])
-
     def register_intent_parser(self, intent_parser):
 
         if hasattr(intent_parser, 'validate') and \

isac/entity_engine.py

@@ -6,9 +6,9 @@
 
 class Entity(object):
 
-    def __init__(self, trie, tokenizer, training, max_tokens=20):
+    def __init__(self, trie, tokenizer, training_data, max_tokens=20):
         self.trie = trie
-        self.training = training
+        self.training_data = training_data
         self.tokenizer = tokenizer
         self.max_tokens = max_tokens
 
@@ -28,8 +28,10 @@ def _sort_and_merge_tags(self, tags):
     def unknown_entities(self, attr, utterance):
 
         pos_tokens = []
-        for _t in self.training:
-            pos_tokens += [t['pos'] for t in _t['tags'] if t['key'] == attr]
+        for intent in self.training_data:
+            for _t in intent:
+                pos_tokens += [t['pos'] for t in _t['tags']
+                               if t['key'] == attr]
 
         # Change to RF classifier or something PLEASE
         most_common_pos = most_common(pos_tokens)

isac/training_engine.py

@@ -0,0 +1,33 @@
+from isac.intent_engine import IntentBuilder
+
+
+class Trainer(object):
+
+    def __init__(self, engine):
+        self.training_data = engine.training_data
+        self.engine = engine
+        self.tokenizer = engine.tokenizer
+        self.tagger = engine.tagger
+
+    def register_entities(self, intent):
+
+        for _data in intent['training']:
+            for entity in _data.get('tags'):
+                self.engine.register_entity(entity['value'], entity['key'])
+
+    def pos_training_data(self):
+
+        for intent in self.training_data:
+            for _t in intent['training']:
+                for tag in _t['tags']:
+                    for pos in self.tokenizer.tagger(_t['text']):
+                        if pos[0] == tag['value']:
+                            tag['pos'] = pos[1]
+
+    def build(self):
+
+        self.pos_training_data()
+        for intent in self.training_data:
+            self.register_entities(intent)
+            self.engine.register_intent(
+                intent['name'], intent['required'], intent['optional'])