Granite Question Classifier: An open-source question classifier - Easily distinguish between general and directed questions

Granite Question Classifier

Developed by cnmoro

A question classifier fine-tuned based on IBM Granite embedding model, used to distinguish between general questions and directed questions

Text Classification

Transformers

Supports Multiple LanguagesOpen Source License:MIT #Problem Classification #RAG Optimization #Multilingual Support

Downloads 22

Release Time : 1/31/2025

Model Overview

This model is a fine-tuned version of ibm-granite/granite-embedding-30m-english, specifically designed to classify question types, distinguishing between general questions (e.g., summary requests) and directed questions (e.g., specific detail queries), to optimize operations in RAG pipelines.

Model Features

High Accuracy Classification

Achieves 94% training accuracy in question classification tasks

Multilingual Support

Supports question classification in English and Portuguese

RAG Pipeline Optimization

Optimizes retrieval-augmented generation workflows by distinguishing question types

Lightweight Model

Built upon an efficient small embedding model

Model Capabilities

Question Classification

Text Classification

Multilingual Processing

RAG Pipeline Integration

Use Cases

Information Retrieval Systems

RAG Pipeline Optimization

Adjusts retrieval strategies based on question types: summary generation for general questions and precise retrieval for directed questions

Improves retrieval efficiency and result relevance

Question Answering Systems

Intelligent Customer Service

Identifies whether user inquiries are general consultations or specific questions, assigning appropriate response strategies

Enhances customer service efficiency and user satisfaction

🚀 Granite Question Classifier

A fine-tuned model for classifying questions into "Directed" or "Generic", enhancing RAG pipeline efficiency.

🚀 Quick Start

This is a fine-tuned version of ibm-granite/granite-embedding-30m-english. Its primary goal is to classify questions into two categories: "Directed" and "Generic". When a question is not directed, it can influence the actions taken in a RAG pipeline. For instance, semantic search might not be directly useful for generic questions like requests for summaries.

During training, this model achieved an accuracy of 94%. It is an upgrade of the previous model: cnmoro/bert-tiny-question-classifier.

✨ Features

Question Classification: Accurately classifies questions into "Directed" or "Generic".
Enhanced RAG Pipeline: Helps optimize actions in a RAG pipeline based on question type.
High Accuracy: Achieved 94% accuracy during training.

📦 Installation

Since this model uses the transformers library, you can install it using the following command:

pip install transformers torch

💻 Usage Examples

Basic Usage

import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification

model_id = "cnmoro/granite-question-classifier"
model = AutoModelForSequenceClassification.from_pretrained(model_id)
tokenizer = AutoTokenizer.from_pretrained(model_id)
model.eval()

def predict_question_category(question):  
    inputs = tokenizer.encode_plus(
        question,
        add_special_tokens=True,
        max_length=512,
        return_tensors="pt",
        truncation=True
    )
    
    input_ids = inputs["input_ids"]
    attention_mask = inputs["attention_mask"]
    
    with torch.no_grad():
        outputs = model(input_ids, attention_mask=attention_mask)
        logits = outputs.logits.squeeze(-1)
        print(logits)
        prediction = (logits > 0).float().item()
    
    # Map prediction to category
    return "directed" if prediction == 1.0 else "generic"

predict_question_category("Qual o resumo do texto?") # generic
predict_question_category("Qual foi a crítica que o autor recebeu do jornal, em relação a sua opinião?") # directed

📚 Documentation

Model Details

Property	Details
Model Type	Fine-tuned version of `ibm-granite/granite-embedding-30m-english`
Training Data	`cnmoro/QuestionClassification-v2`
Pipeline Tag	Text Classification
Base Model	`ibm-granite/granite-embedding-30m-english`
Library Name	`transformers`

Class Mapping

Class 0: Generic
Class 1: Directed

📄 License

This project is licensed under the MIT License.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご