Word Entity Duet:
Learning-to-Rank with Entities in Elasticsearch

The Word Entity Duet project is a Learning to Rank system built on top of open source entity tagging software and a search engine. This project provides scripts for tagging entities in documents and queries using TagMe, and an indexing application for putting those documents and tagged entities into an Elasticsearch index. Additionally, this project provides and entity indexing application which stores entity names, descriptions, and aliases from the Freebase API in an Elasticsearch index. The Feature Generator application creates learning to rank features from tagged queries and generates traditional features such as BM25, Boolean And, and Boolean Or scores; and entity features using the entity IDs as well as the entity features from the Freebase API.

Tutorial Links