🤖 AI Maya Intent Classifier
Lemonade-style conversational AI for insurance
Model Description
This model classifies customer intents in insurance conversations, similar to Lemonade's AI Maya system.
Supported Intents
| Intent | Description | Example |
|---|---|---|
get_quote |
Request policy quote | "How much for car insurance?" |
file_claim |
Start claims process | "I need to file a claim" |
check_status |
Claim/policy status | "What's my claim status?" |
update_policy |
Modify coverage | "Add my spouse to policy" |
cancel_policy |
Policy cancellation | "I want to cancel" |
billing_inquiry |
Payment questions | "When is my payment due?" |
coverage_question |
Coverage details | "Am I covered for floods?" |
general_inquiry |
Other questions | "What's your phone number?" |
Performance Metrics
| Metric | Score |
|---|---|
| Accuracy | 94.2% |
| F1 Score | 93.8% |
| Precision | 94.5% |
| Recall | 93.1% |
Usage
from transformers import pipeline
classifier = pipeline("text-classification", model="gcc-insurance-ml-models/ai-maya-intent-classifier")
result = classifier("I was in a car accident and need to file a claim")
# Output: [{'label': 'file_claim', 'score': 0.97}]
Training Data
Trained on 50,000+ insurance conversation samples covering:
- Policy inquiries
- Claims processing
- Billing questions
- Coverage questions
Lemonade AI Architecture
This model is part of the AI Maya conversational system:
- Intent Classifier (this model) → Understand what customer wants
- Entity Extractor → Extract policy numbers, dates, amounts
- Response Generator → Generate natural language responses
- Sentiment Analyzer → Detect customer emotions
License
Apache 2.0