NLP Pipeline for Support Intelligence

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
from sklearn.preprocessing import LabelEncoder

class SupportTicketClassifier:
    def __init__(self, model_name="bert-base-uncased"):
        self.tokenizer = AutoTokenizer.from_pretrained(model_name)
        self.model = AutoModelForSequenceClassification.from_pretrained(
            model_name,
            num_labels=12
        )
        self.label_encoder = LabelEncoder()

    def preprocess_text(self, text):
        """Clean and tokenize support ticket text"""
        # Remove PII and standardize formatting
        cleaned_text = self.remove_sensitive_data(text)
        return self.tokenizer(
            cleaned_text,
            truncation=True,
            padding='max_length',
            max_length=512,
            return_tensors='pt'
        )

    def classify_intent(self, ticket_text):
        inputs = self.preprocess_text(ticket_text)

        with torch.no_grad():
            outputs = self.model(**inputs)
            predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)

        intent_id = torch.argmax(predictions, dim=-1).item()
        confidence = torch.max(predictions).item()

        return {
            'intent': self.label_encoder.inverse_transform([intent_id])[0],
            'confidence': confidence,
            'urgency_score': self.calculate_urgency(predictions)
        }