Patents in KATI

Why, and how do we analyze patents?

The process of technology foresight using data-based methods is called data driven foresight. With the KATI-system, scientists at Fraunhofer INT have had a system at their disposal for several years that implements this approach in practice. The basis for this is publication data. In the relevant scientific literature, patents are mentioned as another important source of data for technology foresight and innovation management. Our clientele also repeatedly asks for corresponding analyses. But what exactly do we expect from patent analyses? Which questions and use cases should be addressed? And how can patents be integrated into the KATI-system developed at Fraunhofer INT? These are the questions we want to explore in this report.

Why we analyze patents

Patents are a promising source of data because they shed light on technologies in practice: if there are concrete ideas for the application of a technology, there may also be patents. In addition, important insights can be gained about research in companies. More concretely, for example, the market leaders of a technology can be identified, and it can be approximated how much experience a company has in a field. Patent analyses can also be used to make well-founded statements about the maturity of a technology. These and other use cases were identified and sharpened in various projects, workshops and discussions with customers and colleagues. This was flanked by intensive literature research as part of a PhD project – naturally using the in-house KATI-system.

How we handle patent data

In order to perform patent analyses, the data must first be prepared in such a way that it can be processed and analyzed. To this end, the existing architecture of the KATI-system has been extended to include data and analysis options for patents. At first glance, publications and patents may seem quite similar. In both, scientific and technological knowledge is presented, they have an abstract and there is something like an author. However, if you take a closer look, you will notice some decided differences that need to be considered in the data model. For example, in addition to a group of authors (or inventors), patents also have owners, sometimes private individuals, often companies or universities. Furthermore, instead of only one publication date, there are several dates (so-called events), such as the application date or the expected expiration date, which characterize a patent and must be stored in the database. 

The patent data, like the scientific publications, were stored in a graph database. This facilitates the linking of patents with their metadata, such as inventor or filing date, various institutions and even other patents. The ontology, i.e. the underlying net-like schema with the connections of all data among each other, was modified accordingly. In this case, the cleanliness and timeliness of the data must also be considered, because old and unclean data provide less meaningful analyses. To avoid this, a regular update of the data is performed.

How we provide patent analyses

Finally, the user interface was adapted: the search page was tailored to patents and the analyses were adapted, tested and revised with a view to the new data. The latter in particular will continue to be constantly developed and optimized in the coming years, with the aim of getting the best out of the data and creating well-founded insights for early technology intelligence. The basis for this is formed by the various use cases that have already been identified and which are gradually being addressed and implemented. This continues to be done in close exchange with our collegium in the various business units and their customers. In this way, we are adding another important building block to Fraunhofer INT‘s portfolio of methods in a sustainable and forward-looking manner.