Merge pull request #3 from breznak/story-teller

Story teller - randomized start of new sentence

client/linguist.py

@@ -29,11 +29,13 @@
 from nupic.frameworks.opf.modelfactory import ModelFactory
 import model_params
 import re
+import random
+
 
 NUM_REPEATS = 5
 PRINT_EVERY_REPEAT_N = 1
 
-TERMINATORS = ['.','!','?','|']
+TERMINATORS = ['.','!','?','|','\n']
 NUM_SENTENCES = 5 # number for story sentences generated
 STORY_START = "The dog" 
 _QUIT = "QUIT"
@@ -117,19 +119,25 @@ def tellStory(model, startSent, lenght):
     print(c),
     modelInput = {'letter': c}
     result = model.run(modelInput)
-    c=result.inferences['prediction'][0]
-
+    c=result.inferences['prediction'][0] # bug in model randomization of same-probability states 
+    #print "c=",c 
     sentence_len += 1
-    if(sentence_len > 30): #limit, sometimes there's no sentence terminator generated and we'd run forever
-      numSent += 1
-      sentence_len = 0
-
-    if c in TERMINATORS:
+    if (c in TERMINATORS) or (sentence_len > 30): #limit, sometimes there's no sentence terminator generated and we'd run forever
+      # new sentence
       numSent += 1
       sentence_len = 0
       print(' \n')
+      c = _generateRandomStartingLetter()
+
+
+###################################################################
+def _generateRandomStartingLetter(): 
+  """generates a random starting letter for sequence after the RESET.
+     TODO: add probability based distribution later"""
+  return chr(random.randint(65,90)) # A-Z == chr(65)-chr(90)
 
 
+###################################################################
 if __name__ == "__main__":
   if len(sys.argv) > 1:
     datapath = sys.argv[1]

client/model_params.py

@@ -228,7 +228,7 @@
             # elements to append to the end of a learned sequence at a time.
             # Smaller values are better for datasets with short sequences,
             # higher values are better for datasets with long sequences.
-            'pamLength': 5,
+            'pamLength': 6,
         },
 
         'clParams': {
@@ -240,14 +240,14 @@
 
             # This controls how fast the classifier learns/forgets. Higher values
             # make it adapt faster and forget older patterns faster.
-            'alpha': 0.01080959960713530062,
+            'alpha': 0.001080959960713530062,
 
             # This is set after the call to updateConfigFromSubConfig and is
             # computed from the aggregationInfo and predictAheadTime.
             #'steps': '1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16',
             'steps': '1,2,3,4,5',
         },
 
-        'trainSPNetOnlyIfRequested': True,
+        'trainSPNetOnlyIfRequested': False,
     },
 }