| language: en | |
| license: mit | |
| pipeline_tag: text-classification | |
| tags: | |
| - nlp | |
| - transformers | |
| - roberta | |
| - peft | |
| - lora | |
| - multitask-learning | |
| - customer-support | |
| # TicketIQ-MultiTask | |
| TicketIQ is a multi-task NLP model for automated customer support ticket triage. | |
| A single RoBERTa encoder fine-tuned with LoRA predicts: | |
| - Ticket Category | |
| - Ticket Priority | |
| - Customer Sentiment | |
| All predictions are produced in a single forward pass. | |
| --- | |
| ## Model Overview | |
| | Task | Example Labels | | |
| |--------|--------| | |
| | Category | account, billing, technical, shipping | | |
| | Priority | low, medium, high, critical | | |
| | Sentiment | positive, neutral, negative | | |
| Architecture: | |
| ```text | |
| RoBERTa Base | |
| │ | |
| ▼ | |
| Shared Encoder | |
| │ | |
| ┌────┼────┐ | |
| ▼ ▼ ▼ | |
| Category | |
| Priority | |
| Sentiment | |
| Heads | |