You're viewing version 3.5 of the OpenSearch documentation. This version is no longer maintained. For the latest version, see the current documentation. For information about OpenSearch version maintenance, see Release Schedule and Maintenance Policy.
Learning to Rank
The Learning to Rank plugin for OpenSearch enables you to use machine learning (ML) and behavioral data to fine-tune the relevance of documents. It uses models from the XGBoost and RankLib libraries. These models rescore the search results, considering query-dependent features such as click-through data or field matches, which can further improve relevance.
The term learning to rank is abbreviated as LTR throughout the OpenSearch documentation when the term is used in a general sense. For the plugin developer documentation, see opensearch-learning-to-rank-base.
Getting started
The following resources can help you get started:
- If you are new to LTR, start with the ML ranking core concepts documentation.
- For a quick introduction, see the demo in hello-ltr.
- If you are familiar with LTR, start with the Scope of the plugin documentation.
Core concepts and setup
Before implementing LTR, familiarize yourself with the foundational concepts and architecture:
- ML ranking core concepts: Understand the fundamental concepts behind Learning to Rank.
- Scope of the plugin: Learn how LTR integrates with your OpenSearch infrastructure.
Feature engineering and model development
Create and train your ranking models using the following workflow:
- Feature engineering: Design effective features for your ranking models.
- Working with features: Create and manage feature sets.
- Logging feature scores: Collect feature data for model training.
- Uploading trained models: Build and train your ranking models.
Deployment and advanced topics
Once your models are trained, deploy them in production and explore advanced features:
- Optimizing search with LTR: Deploy models in production search.
- Advanced functionality: Explore advanced LTR features and techniques.
- Common issues: Common questions and troubleshooting.