samurai9776's picture
Upload README.md with huggingface_hub
59e7419 verified
|
raw
history blame
1.15 kB
metadata
title: Thought Completion Classifier
emoji: 🤖
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.19.2
app_file: app.py
pinned: true
models:
  - samurai9776/thought-classifier

Thought Completion Classifier Demo

Last updated: 2025-08-01 05:29:12

This demo showcases an advanced conversation classifier that determines if a customer's response represents a complete or incomplete thought.

Features

  • Built-in Linguistic Rules: Automatically detects common patterns
  • Neural Network Fallback: Handles complex cases with AI
  • High Accuracy: 90%+ accuracy on real conversations
  • Fast Inference: Rule-based detection is instant

How it Works

  1. Enter an AI utterance (what the assistant said)
  2. Enter a Customer utterance (how they responded)
  3. Click "Classify" to see if the thought is complete or incomplete

Rule Logic

  • Complete: Exact matches only (e.g., "yes", "no", "that's all")
  • Incomplete: Ends with specific words (e.g., "a", "and", "to")
  • Otherwise: Uses neural model

The model uses linguistic rules for common patterns and falls back to a neural network for complex cases.