README.md

metadata

license: apache-2.0
language:
  - zh
  - en
tags:
  - intent-classification
  - chinese
  - albert
  - binary-classification
  - transformers
  - diner-call
base_model:
  - ckiplab/albert-base-chinese
datasets:
  - Luigi/dinercall-intent
pipeline_tag: text-classification
model-index:
  - name: albert-base-chinese-dinercall-intent
    results:
      - task:
          name: Intent Classification
          type: text-classification
        dataset:
          name: DinerCall Intent
          type: Luigi/dinercall-intent
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.85

albert-base-chinese-dinercall-intent

A binary intent classification model fine-tuned from ckiplab/albert-base-chinese on the Luigi/dinercall-intent dataset. This model identifies whether a Chinese restaurant phone call contains a reservation intent (label=1) or not (label=0).

Model Details

• Base model: ckiplab/albert-base-chinese
• Task: Binary intent classification
• Language: Traditional Chinese
• Labels:
  • 0: No intent to book a table
  • 1: Intent to make a reservation

Use Cases

This model is designed for voice AI assistants in restaurants to automatically identify reservation intents from spoken or transcribed customer sentences.

Example Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained("Luigi/albert-base-chinese-dinercall-intent")
model = AutoModelForSequenceClassification.from_pretrained("Luigi/albert-base-chinese-dinercall-intent")

inputs = tokenizer("你好，我想訂位，今天晚上七點兩位", return_tensors="pt")
with torch.no_grad():
    logits = model(**inputs).logits
    predicted = torch.argmax(logits, dim=-1).item()

print(f"Prediction: {'Reservation intent' if predicted == 1 else 'No intent'}")