| language: en | |
| tags: | |
| - text-classification | |
| - conversational-ai | |
| - thought-completion | |
| widget: | |
| - text: "Great. Would you like anything else? [SEP] Picking up on app order." | |
| example_title: "Complete thought" | |
| - text: "Your total is $15.99 [SEP] Actually, let me add" | |
| example_title: "Incomplete thought" | |
| # Thought Completion Classifier | |
| This model classifies whether a conversation between an AI assistant and a customer represents a complete or incomplete thought. | |
| ## Model Description | |
| - **Model Type:** DistilBERT fine-tuned for sequence classification | |
| - **Task:** Binary classification (Complete/Incomplete thought) | |
| - **Training Data:** 900+ labeled conversation examples | |
| - **Input Format:** "AI_Utterance [SEP] CX_Utterance" | |
| ## Usage | |
| ```python | |
| from transformers import pipeline | |
| # Load the model | |
| classifier = pipeline("text-classification", model="samurai9776/thought-classifier") | |
| # Example usage | |
| result = classifier("What else can I get you? [SEP] That's all for now") | |
| print(result) | |
| # Output: [{'label': 'COMPLETE', 'score': 0.95}] | |
| ``` | |
| ## Training Details | |
| - Base Model: distilbert-base-uncased | |
| - Training epochs: 3 | |
| - Batch size: 16 | |
| - Learning rate: 2e-5 | |
| ## Limitations | |
| - Trained on English conversations only | |
| - Optimized for customer service/order-taking scenarios | |