مقاله سال ۲۰۱۶

A lot of corporate technology managers and scientists in research and political institutions seek to understand typical evolutionary patterns of technological fields. For instance, they might wish to know which topics are relevant for the respective technological field, which are the emerging topics, which traditional topics have been deepened in the course of time and which have been abandoned? With respect to many, though by no means all, technological fields, patent analyses may help to answer such questions. They have been used successfully and extensively in many cases, and comprise different techniques, such as co-classification analysis (see as examples Choi et al., 2007; Chang et al., 2009; Dereli and Durmusoglu, 2009) or citation analysis (Tseng et al., 2011; Frietsch, 2007; Kuusi and Meyer, 2007; de Souza Carvalho et al., 2009; Lee et al., 2009, 2012).

A multitude of techniques for patent analysis makes use of the so-called meta-data of patents. Meta-data are defined by patent laws like the U.S. Code Title 35 and comprise information on applicants, inventors, classifications (international patent classification [IPC], current patent classification [CPC], in some cases national classifications like the US patent classification [USPC]), application and granting dates, cited patents and other literature (Ernst, 2003; Lee et al., 2011). Valuable answers to some of the questions mentioned above can be obtained by such analyses; techniques like activity analysis, co-classification activity analysis and citation network analysis may provide answers regarding different technical aspects and the development of topics in the technological field over time. The answers to some questions are not quite perfect yet, and there is still potential for improvement. Especially exploiting the information contained in the full-text of patents (instead or in addition to meta-data) by means of text mining technologies, as suggested by Yoon and Kim (2011) as well as by Moehrle and Gerken (2012) and Gerken and Moehrle (2012), may provide researchers with deeper insights. Text mining offers the opportunity to establish semantic similarity measures between documents and in doing so provides an alternative or an addition to the well known citation analysis.

In this paper we concentrate on these text mining technologies and suggest so-called patent lanes which we define as the deployment of patent clusters over time. The idea of patent lanes is related to the timeline visualization of the development of technological clusters (Shibata et al., 2010), but uses disaggregated information instead of clusters. The paper is organized as follows: In the next section we briefly explain how to measure semantic similarities between patents, as this constitutes the foundation of our method. In order to underpin our methodical contribution, we compare semantic similarities and citations as basic elements that establish links between patents, and show their interrelation in analyses. As patent lanes may be configured in different ways, we discuss the most important design decisions. A case study which focuses on carbon fiber reinforcements and the utilization thereof in bicycle technology, serves to illustrate the use of patent lanes and the interpretation of results. We compare our method with methods characterized by rolling clustering to identify criteria for the usefulness of its application. Some concluding remarks will highlight implications as well as limitations of our method.