Improvements in Recall and Precision in Wolters Kluwer Spain Legal Search Engine

Abstract

In this paper we describe the search technology in production in Wolters Kluwer Spain for the legal research market. This technology improves the “Google like” experience by increasing both the total number of retrieved documents (recall) and the quality of the very best ones (precision) while maintaining the ease of entering a natural language query. We propose a hybrid approach, both in the working methodology and in the codification of the legal knowledge -subject matter expert and librarian-, through new layers of semantic analysis and algorithms. We improve the traditional tf-idf vector space model by creating a mixed document indexing schema of terms and concepts as well as a proprietary ranking algorithm trained by a hybrid genetic algorithm. These calculations also improve the quality of keyword-in-context. Keywords Legal Knowledge Representation Information Retrieval Semantic Indexation Hybrid Methodologies Genetic Algorithms Machine Learning