Information Technology-Based Patent Retrieval Models

Abstract

This chapter presents information technology ( ) based patent retrieval models. It first compares and contrasts information retrieval ( ) with patent retrieval, and highlights their key differences. For instance, IR can be considered as a precision-oriented retrieval, whereas patent retrieval can be considered as a recall-oriented retrieval. The chapter then describes the boolean retrieval model, which was designed for IR but can be used for patent retrieval. To facilitate effective patent retrieval, a basic patent retrieval model is presented. With this model, representative keyword terms are extracted from the user query and are ranked according to their importance so that top-$k$ relevant patents can be retrieved with irrelevant patents eliminated. Moreover, the chapter also presents some enhancements and extensions to the basic patent retrieval model, which include incorporation of relevance feedback, estimation of the importance of keyword terms, text preprocessing of patent documents, and handling of patent category frequency. In addition, two dynamic patent retrieval models are also described. These two models perform interactive patent retrieval via dispersion or accumulation to dynamically rank the patents. Experimental results with real-life datasets show that the models presented in this chapter outperformed many conventional search systems with respect to time and cost. While this chapter focuses on the theoretical aspects of IT based patent retrieval models which are of interest to IT specialists, practical illustrative examples in the chapter demonstrate the empirical aspects of patent retrieval models which are helpful to IT practitioners.