Multiple Retrieval Models and Regression Models for Prior Art Search

TL;DR

This paper introduces PATATRAS, a patent retrieval system combining multiple models, regression-based result merging, and metadata utilization to improve prior art search across three languages.

Contribution

The paper presents a novel multi-model, regression-based approach that integrates patent metadata and citation data for enhanced prior art search effectiveness.

Findings

Multiple retrieval models produce diverse ranked results.

Regression merging improves overall search accuracy.

Metadata and citation data enhance initial set creation and re-ranking.

Abstract

This paper presents the system called PATATRAS (PATent and Article Tracking, Retrieval and AnalysiS) realized for the IP track of CLEF 2009. Our approach presents three main characteristics: 1. The usage of multiple retrieval models (KL, Okapi) and term index definitions (lemma, phrase, concept) for the three languages considered in the present track (English, French, German) producing ten different sets of ranked results. 2. The merging of the different results based on multiple regression models using an additional validation set created from the patent collection. 3. The exploitation of patent metadata and of the citation structures for creating restricted initial working sets of patents and for producing a final re-ranking regression model. As we exploit specific metadata of the patent documents and the citation relations only at the creation of initial working sets and during the final post ranking step, our architecture remains generic and easy to extend.