Poprawki - usuwamy convlaba

src/main.py

@@ -11,9 +11,6 @@ monitor = DialogStateMonitor() # DSM
 dialog_policy = DialogPolicy() # DP
 language_generation = NaturalLanguageGeneration(templates) # NLG
 
-agent = PipelineAgent(nlu=nlu, dst=monitor, policy=None, nlg=language_generation, name='sys')
-resp = agent.response("Dzień dobry")
-print(resp)
 # Main loop
 dial_num = 0
 print("CTRL+C aby zakończyć program.")
@@ -35,8 +32,7 @@ while True:
         system_action = dialog_policy.predict(monitor)
 
         # NLG
-        act, slots = parse_frame(frame)
-        response = language_generation.generate(act, slots)
+        response = language_generation.generate(frame)
         print(response)
 
         if frame.act == "bye":

src/service/dialog_policy.py

@@ -1,6 +1,6 @@
 from collections import defaultdict
 from model.frame import Frame
-from src.model.slot import Slot
+from model.slot import Slot
 
 class DialogPolicy:
     

src/service/dialog_state_monitor.py

@@ -1,4 +1,4 @@
-from src.model.frame import Frame
+from model.frame import Frame
 import copy
 import json
 
@@ -9,7 +9,7 @@ def normalize(value):
 
 
 class DialogStateMonitor:
-    def __init__(self, initial_state_file: str = '../attributes.json'):
+    def __init__(self, initial_state_file: str = 'attributes.json'):
         with open(initial_state_file) as file:
             constants = json.load(file)
         self.__initial_state = dict(belief_state={

src/service/natural_languag_understanding.py

@@ -1,8 +1,7 @@
-from convlab.nlu.nlu import NLU
 from flair.models import SequenceTagger
 from utils.nlu_utils import predict_single, predict_and_annotate
 from model.frame import Frame, Slot
-
+import random
 """
 ACTS:
     inform/order
@@ -42,7 +41,7 @@ SLOTS:
     sauce
 """
 
-class NaturalLanguageUnderstanding(NLU):
+class NaturalLanguageUnderstanding():
     def __init__(self):
         print("\n========================================================")
         print("Models are loading, it may take a moment, please wait...")
@@ -86,8 +85,15 @@ class NaturalLanguageUnderstanding(NLU):
 
         return slots
 
-    def predict(self, text: str, context: list):
+    def predict(self, text: str):
         act = self.__predict_intention(text)
         slots = self.__predict_slot(text)
         frame = Frame(source = 'user', act = act, slots = slots)
+        # uncomment to quickly mock the response
+        # frames = [
+        #     Frame(source="user", act = "inform/order", slots=[Slot(name="pizza", value="barcelona")]),
+        #     Frame(source="user", act = "welcomemsg", slots=[]),
+        #     Frame(source="user", act = "request/menu", slots=[]),
+        # ]
+        # return random.choice(frames)
         return frame

src/service/natural_language_generation.py

@@ -1,26 +1,26 @@
-import re
-from service.template_selector import select_template
-import random
-from convlab.nlg.nlg import NLG
-
-# from service.templates import templates
-
-def parse_frame(frame):
-    if not hasattr(frame, 'act') or not hasattr(frame, 'slots'):
-        raise TypeError("Expected a Frame object with 'act' and 'slots' attributes.")
-    
-    act = frame.act
-    slots = [{"name": slot.name, "value": slot.value} for slot in frame.slots]
-    
-    return act, slots
-
-class NaturalLanguageGeneration(NLG):
-    def __init__(self, templates):
-        self.templates = templates
-
-    def generate(self, act, slots):
-        template = select_template(act, slots)
-        if template == "default/template":
-            template = random.choice(self.templates["default/template"])
-        slot_dict = {slot['name']: slot['value'] for slot in slots}
+import re
+from service.template_selector import select_template
+import random
+
+# from service.templates import templates
+
+def parse_frame(frame):
+    if not hasattr(frame, 'act') or not hasattr(frame, 'slots'):
+        raise TypeError("Expected a Frame object with 'act' and 'slots' attributes.")
+    
+    act = frame.act
+    slots = [{"name": slot.name, "value": slot.value} for slot in frame.slots]
+    
+    return act, slots
+
+class NaturalLanguageGeneration():
+    def __init__(self, templates):
+        self.templates = templates
+
+    def generate(self, frame):
+        act, slots = parse_frame(frame)
+        template = select_template(act, slots)
+        if template == "default/template":
+            template = random.choice(self.templates["default/template"])
+        slot_dict = {slot['name']: slot['value'] for slot in slots}
         return template.format(**slot_dict)