Editor’s note: Mohammad Soltanieh-ha, PhD is a speaker for ODSC East 2025 this May 13th-15th. Be sure to check out his talk, “Hybrid Text Classification: Labeling with LLMs and Dense Neural Networks,” there!

In many machine learning projects, labeling text data is a critical yet resource-intensive process. Relying solely on premium Large Language Models (LLMs) for labeling can quickly become expensive and slow, especially when scaling up. Recognizing this challenge, our upcoming workshop at ODSC East 2025 introduces a hybrid approach that leverages the strengths of both premium LLM APIs and open-source tools such as fully-connected neural networks. 

A Hybrid Approach to Text Labeling 

The session begins by exploring how premium LLMs, such as those accessible via the OpenAI API, can jump-start the labeling process. These models provide high-quality initial labels, which form the foundation of your dataset. However, as you’ll learn, integrating these labels with open-source tools allows for greater flexibility and cost savings. 

Participants will learn how to generate vector embeddings using open-source models like Sentence Transformers (specifically, the ‘all-MiniLM-L6-v2’ model). These embeddings not only capture the nuanced semantic relationships within text data but also serve as the key input for training robust neural network classifiers. By combining the strengths of both premium and open-source solutions, you can achieve rapid, cost-effective text classification without sacrificing accuracy. 

Workshop Session Outline 

Lesson 1: Introduction to LLM-Assisted Labeling 

• Overview: Understand the benefits and limitations of using premium LLMs for text labeling. 
• Hands-on Exercise: Use OpenAI APIs to label a small dataset and analyze the output, gaining insights into how automated labeling can streamline your workflow. 

Lesson 2: Vector Embeddings for Text Representation 

• Concepts Covered: Dive into the world of vector embeddings and learn why they are essential for effective text classification.
• Hands-on Exercise: Utilize open-source models in Google Colab to generate embeddings from your labeled data and visualize these embeddings to better understand their distribution. 

Lesson 3: Training a Neural Network Classifier 

• Deep Dive: Explore the architecture and workflow of neural networks tailored for text classification tasks. 
• Hands-on Exercise: Train a neural network classifier using Keras on the embeddings generated earlier. Evaluate its performance and compare the results with those obtained using only premium LLMs, highlighting the benefits of a hybrid approach. 

Tools and Learning Objectives 

During the workshop, you’ll gain practical skills in: 

• Using the OpenAI API: Learn how to effectively employ premium LLMs for initial text labeling. 
• Google Colab: Utilize this platform for running experiments and prototyping your text analytics workflows. 
• Sentence Transformers: Extract high-quality vector embeddings using models like ‘all-MiniLM-L6-v2’. 
• Keras: Build, train, and evaluate neural network classifiers that deliver fast and accurate text classification. 

This session is designed for professionals with some background in Python and neural networks, targeting an intermediate to advanced audience. By the end of the workshop, you’ll be equipped with the knowledge to create cost-effective, scalable, and high-performing text classification systems that blend the best of both premium and open-source technologies. 

About the Author/ODSC East 2025 Speaker on Hybrid Text Classification: