162 lines
5.5 KiB
Python
162 lines
5.5 KiB
Python
import spacy
|
|
|
|
nlp = spacy.load("pl_core_news_md")
|
|
|
|
class DialogManager:
|
|
def __init__(self, nlu_module, dst_module):
|
|
self.nlu_module = nlu_module
|
|
self.dst_module = dst_module
|
|
|
|
def start_dialog(self):
|
|
self.dst_module.update_state([]) # Zerowanie stanu dialogowego
|
|
|
|
while True:
|
|
user_input = input("Użytkownik: ")
|
|
|
|
# Wykorzystanie modułu NLU do ekstrakcji aktywności i slotów
|
|
acts = self.nlu_module.extract_acts_and_slots(user_input)
|
|
|
|
# Aktualizacja stanu dialogowego za pomocą modułu DST
|
|
self.dst_module.update_state(acts)
|
|
dialog_state = self.dst_module.get_state()
|
|
|
|
# Logika dialogu
|
|
if not acts:
|
|
response = "Przepraszam, nie rozumiem. Czym mogę Ci pomóc?"
|
|
elif "hello" in acts[0].act_type:
|
|
response = "Witaj! W czym mogę Ci pomóc?"
|
|
elif "inform" in acts[0].act_type:
|
|
response = self.generate_response(dialog_state)
|
|
elif "bye" in acts[0].act_type:
|
|
response = "Dziękuję za rozmowę. Miłego dnia!"
|
|
break
|
|
else:
|
|
response = "Nie rozumiem. Czym mogę Ci pomóc?"
|
|
|
|
print("Agent:", response)
|
|
|
|
def generate_response(self, dialog_state):
|
|
# Logika generowania odpowiedzi na podstawie stanu dialogowego
|
|
# Możesz dostosować tę logikę do swoich potrzeb
|
|
# Przykład: generowanie odpowiedzi na podstawie aktualnego stanu dialogowego
|
|
response = "Rozumiem, potrzebujesz"
|
|
|
|
if "product type" in dialog_state and "product" in dialog_state:
|
|
product_type = dialog_state["product type"]
|
|
product = dialog_state["product"]
|
|
response += f" {product} z kategorii {product_type}"
|
|
elif "product type" in dialog_state:
|
|
product_type = dialog_state["product type"]
|
|
response += f" produkty z kategorii {product_type}"
|
|
else:
|
|
response += " informacji o twoich potrzebach"
|
|
|
|
response += ". Jak mogę Ci jeszcze pomóc?"
|
|
|
|
return response
|
|
|
|
class DialogStateTracker:
|
|
def __init__(self):
|
|
self.dialog_state = {}
|
|
|
|
def update_state(self, acts):
|
|
for act in acts:
|
|
if act.act_type == "hello":
|
|
self.dialog_state = {}
|
|
elif act.act_type == "bye":
|
|
self.dialog_state = {}
|
|
elif act.act_type == "inform":
|
|
slots = act.slots
|
|
for slot, value in slots.items():
|
|
self.dialog_state[slot] = value
|
|
|
|
def get_state(self):
|
|
return self.dialog_state
|
|
|
|
|
|
|
|
|
|
|
|
product_type_rules = {
|
|
"pieczywo": ["chleb", "bułka", "rogalik", "bagietka"],
|
|
"owoce": ["jabłko", "banan", "gruszka", "pomarańcza"],
|
|
"warzywa": ["marchew", "ziemniak", "cebula", "pomidor"],
|
|
"mięso": ["kurczak", "wołowina", "wieprzowina", "indyk"],
|
|
"produkty mrożone": ["lody", "frytki", "pierogi mrożone", "nuggetsy"],
|
|
"słodycze": ["czekolada", "ciastko", "lizak", "guma do żucia"],
|
|
"przyprawy": ["sól", "pieprz", "oregano", "cynamon"],
|
|
"napoje": ["woda", "sok", "herbata", "kawa"],
|
|
"napoje alkoholowe": ["piwo", "wino", "wódka", "whisky"],
|
|
"higiena": ["pasta do zębów", "mydło", "szampon", "papier toaletowy"],
|
|
"chemia gospodarcza": ["płyn do naczyń", "proszek do prania", "odświeżacz powietrza"],
|
|
"inne": ["długopis", "baterie", "śrubokręt", "nożyczki"],
|
|
"nabiał": ["mleko"]
|
|
}
|
|
|
|
class DialogAct:
|
|
def __init__(self, act_type, slots=None):
|
|
self.act_type = act_type
|
|
self.slots = slots if slots else {}
|
|
|
|
class NLU:
|
|
def __init__(self):
|
|
pass
|
|
|
|
def extract_acts_and_slots(text):
|
|
doc = nlp(text)
|
|
acts = []
|
|
|
|
for token in doc:
|
|
if token.lower_ == "cześć":
|
|
acts.append(DialogAct("hello"))
|
|
elif token.lower_ == "do widzenia":
|
|
acts.append(DialogAct("bye"))
|
|
elif token.lower_ == "dziękuję":
|
|
acts.append(DialogAct("thankyou"))
|
|
elif token.lower_ == "proszę":
|
|
acts.append(DialogAct("request"))
|
|
elif token.lower_ == "powtórz":
|
|
acts.append(DialogAct("repeat"))
|
|
elif token.lower_ == "reset":
|
|
acts.append(DialogAct("restart"))
|
|
elif token.lower_ in ["tak", "oczywiście"]:
|
|
acts.append(DialogAct("affirm"))
|
|
elif token.lower_ in ["nie", "nie chcę"]:
|
|
acts.append(DialogAct("deny"))
|
|
elif token.pos_ == "NOUN":
|
|
product_type, product = self.find_product_type(token.lemma_)
|
|
if product_type and product:
|
|
act = DialogAct("inform", {"product type": product_type, "product": product})
|
|
acts.append(act)
|
|
|
|
return acts
|
|
|
|
def find_product_type(product):
|
|
for product_type, products in product_type_rules.items():
|
|
if product in products:
|
|
return product_type, product
|
|
|
|
return None, None
|
|
|
|
text = "Cześć, chciałbym kupić mleko, sok oraz chleb"
|
|
nlu = NLU()
|
|
acts = nlu.extract_acts_and_slots(text)
|
|
for act in acts:
|
|
print(f"Type: {act.act_type}")
|
|
print(f"Slots: {act.slots}")
|
|
print()
|
|
|
|
|
|
dst = DialogStateTracker()
|
|
|
|
text = "Cześć, chciałbym kupić mleko, sok oraz chleb"
|
|
acts = nlu.extract_acts_and_slots(text)
|
|
|
|
|
|
dst.update_state(acts)
|
|
|
|
|
|
dialog_state = dst.get_state()
|
|
|
|
print("Dialog State:")
|
|
print(dialog_state) |