Combinational inventions

"New technologies are never created from nothing. They are constructed-put together-from components that previously exist", according to Arthur and Polak in 2006. This article explores how to identify combinatorial inventions that originate from combining two technologies in a new way. The importance of combinatorial inventions is confirmed by anecdotal evidence from examiners, who report that groundbreaking discoveries of unknown technologies are a rather rare occurrence. In their 2015 article entitled "Invention as a combinatorial process: evidence from US patents", authors Hyejin,Youn et al. describe how analysing the co-assignments of classification symbols can be used to show the importance of technological combinations for innovation. They analyse the distance between classification codes to determine the novelty of such combinations and how "the generation of novel technological combinations engenders a practically infinite space of technological configurations."

In line with this concept, the EPO's tools Global Patent Index (GPI) and PATSTAT can be used to analyse the CPC co-assignments of patent publications in the result set of any search with a view to identifying new technological trends that have not yet resulted in new classification symbols. The underlying idea is that changes over time in combinatorial patterns are used to extract novel combinations from the background of typical combinations in a specific area. These novel combinations may have the potential to spur further inventions and become trendsetters for new technologies.

Figure 1 shows the combinations of CPC assignments in relation to inventions for traffic control of road vehicles (G08G1/00). In this chart, classic combinations with data processing classified under G06Q are predominant and create so much noise that new trends with low filing numbers are difficult to spot.

Figure 1 CPC co-assignments in traffic control (G08G1/00)
(Click image to enlarge) – Figure 1 CPC co-assignments in traffic control (G08G1/00)

Figure 2, however, plots the change over time in CPC co-assignments of traffic control systems (G08G1/00) with other areas, highlighting combinations that increase in importance (blue circles) as well as those that decrease (red circles). This chart clearly shows that in recent times data from transport infrastructure is more often co-assigned with information from moving vehicles, e.g. navigation systems in G05D1/00 and even from autonomous vehicles B60W60/00. Such analyses can help to identify technological trends that have not yet translated into high filing numbers and, therefore, provide early insight into future trends.

 Figure 2 Change over time of CPC co-assignments in traffic control (G08G1/00)

(Click image to enlarge) – Figure 2 Change over time of CPC co-assignments in traffic control (G08G1/00)

The EPO's public patent information tools currently allocate CPC codes at patent‑family level, thereby classifying the entire document. Newer approaches aim to advance the concept of analysing classification assignments in more detail by annotating the classification codes not only at document level but also at paragraph level to guide the examiner to the right paragraph. This also improves combinatorial searches since it makes it possible to investigate whether the co‑assignments are close together (e.g. in the same paragraph and thus refer to the same concept) or far apart and therefore potentially unrelated.

As an example, Figure 3 shows the annotation of classification codes for each section in the document. Comparing similar sets of combinations of codes indicates a potential relationship between the technical concepts in the different documents.

 Figure 3 Using AI to assign CPC codes at paragraph level to identify similar close combinations

(Click image to enlarge) – Figure 3 Using AI to assign CPC codes at paragraph level to identify similar close combinations

Combinatorial inventions as described in this article can already be explored with EPO tools to identify novel combinations for trendspotting. In the future, it may be possible to use AI to analyse the text and even to propose new classification codes for concepts that are not yet covered by existing classification codes.

Quick Navigation