|
import os |
|
from openai import OpenAI |
|
from models.query import Query |
|
import re |
|
from transformers import pipeline |
|
from services.utils import clean_text, encode_and_normalize |
|
|
|
intent_recognizer = pipeline( |
|
"zero-shot-classification", model="facebook/bart-large-mnli") |
|
|
|
ner_recognizer = pipeline( |
|
'ner', model='dbmdz/bert-large-cased-finetuned-conll03-english') |
|
|
|
openai_key = os.environ.get("OPENAI_KEY") |
|
openai_client = OpenAI(api_key=openai_key) |
|
|
|
|
|
patterns = { |
|
'Product': r'\b(iphone|samsung|macbook|ps5|galaxy|pixel|shoes|shampoo|cellphone|smartphone|tablet|laptop|headphones|console|tv|camera)\b', |
|
'Brand': r'\b(apple|samsung|google|sony|microsoft|dell|hp|lenovo|asus|nintendo|canon|nikon)\b', |
|
'Category': r'\b(laptops|dresses|phones|electronics|clothing|footwear|accessories|home appliances|furniture)\b', |
|
'Color': r'\b(red|black|yellow|blue|green|white|grey|pink|purple|orange|brown)\b', |
|
'Price Range': r'\b(under \$?\d+|below \$?\d+|less than \$?\d+|more than \$?\d+|above \$?\d+|between \$?\d+ and \$?\d+)\b', |
|
'Quantity': r'\b(\d+ bottles|\d+ items|\d+ pieces|\d+ units|\d+)\b', |
|
'Order Number': r'\bB-\d+\b', |
|
'Issue': r'\b(account help|payment issue|order problem|shipping delay|return request|product complaint|technical support)\b' |
|
} |
|
|
|
INTENTS = [ |
|
"search for products", |
|
"order management", |
|
"checkout", |
|
"customer support", |
|
"account management" |
|
] |
|
|
|
|
|
def recognize_intent(text): |
|
cleaned_text = clean_text(text) |
|
intent = intent_recognizer(cleaned_text, INTENTS) |
|
return intent['labels'][0] |
|
|
|
|
|
def recognize_entities(text): |
|
cleaned_text = clean_text(text) |
|
entities_from_ner = ner_recognizer(cleaned_text) |
|
entities_from_re = {entity: re.findall(pattern, text.lower( |
|
)) for entity, pattern in patterns.items() if re.findall(pattern, text.lower())} |
|
|
|
return entities_from_re |
|
|
|
|
|
def extract_keywords(text): |
|
cleaned_text = clean_text(text) |
|
return cleaned_text.split() |
|
|
|
|
|
def generate_response(query: Query, context_from_elasticsearch): |
|
prompt = query.history |
|
prompt.append({"role": "assistant", "content": str(context_from_elasticsearch)}) |
|
|
|
print(prompt) |
|
|
|
response = openai_client.chat.completions.create( |
|
model="gpt-3.5-turbo", |
|
messages=prompt |
|
) |
|
|
|
return response.choices[0].message.content |
|
|
|
|
