Javier Alonso on lead optimisation at Idealista

Optimizing Lead Classification at Idealista Using Argilla and GPT-4

In this meetup, join David and Javi from Idealista to explore their recent project on optimizing lead classification. Leveraging Argilla and GPT-4, they tackle the issue of filtering thousands of user messages to ensure only relevant and appropriate communications reach real estate agents. The session covers the entire journey from initial problem identification, data extraction, auto labeling using GPT-4, to human verification and model training. Gain insights into the importance of accurate data, innovative labeling strategies, and the technical tools that enabled the success of this project. Tune in to learn how Idealista efficiently manages lead categorization and enhances its operational workflow.

00:00 Introduction and Guest Welcome
01:40 Overview of Idealista
03:47 Understanding the Problem: Lead Classification
10:31 Data Collection and Preparation
17:00 Auto Labeling with Large Language Models
24:42 Introducing Auto Labels with Large Language Models
24:51 Exploring Argilla's Suggestions Feature
26:28 Batch Labeling for Efficient Processing
27:41 Iterative Record Increase and Model Training
30:11 Fine-Tuning Transformers for Text Classification
33:21 Conclusions and Key Takeaways
38:34 Q&A Session: Language Models and Project Insights
48:48 Final Remarks and Future Plans