A machine learning view to Patent portfolios of selected Telecommunication companies
Could we look at what happens when companies like Nokia and Alcatel-Lucent merge their patent portfolios? In a recent article I published with my co-authors Hannes Toivanen and Marko Seppänen in the journal Technological Forecasting and Social Change, we show a method of using patent data to structure company knowledge depth and breadth enabling a comparative analysis of competing companies. An approach that can be handy in Merger & Acquisition, Technology & Knowledge management, and making investment decisions.
So, would we be interested in technology management, technological forecasting or day-to-day patent monitoring, we do need efficient tools to quantify the knowledge resources of a company – patents often being the only proxy we have at our disposal. Patents are interesting not just due to their legal aspects, but also as it gives an window into what companies have knowledge on. For example, in the case of Nokia and Alcatel-Lucent, is Nokia technology vise buying something that they already have, something obsolete or something new (or new to Nokia).
Existing methods looking at patent information usually involves keyword searches, using patent classifications and managing excel-sheets – an effort that can be
time consuming and might not provide consistent result. Using natural language processing we can move beyond existing human-effort based methods towards a more consistent and dynamic view of knowledge landscapes.
In the article, we took an interesting sample of the telecommunication industry, namely Alcatel-Lucent, Apple, Google, Huawei, Microsoft, Nokia and Samsung Electronics. This an interesting sample as it provides a rich data set where companies have changed their focus, been acquired and altogether the companies have a very different knowledge profile. Using unsupervised learning to analyse the patents of the companies we quantified their knowledge profiles, gaining a holistic view of their knowledge assets. We showed differences between software-oriented companies (such as Google and Microsoft) and technology-driven firms (such as Nokia or Huawei), underlining that they have a different focus in their knowledge base. The results of our analysis is in the Figure below and also provided as a dynamic graph on my website.
For technology managers the temporal dimension of the analysis is especially valuable, as we could for example show the systemic transition of the telecommunication industry towards a software-driven knowledge frame, but also detecting those hardware-related areas that are growing against the overall trend. This allows for dynamic monitoring, but also forecasting innovation pathways.
In our experience, the method has several inherent advantages, of which most important ones are the versatility and agility of analysis: Any large data set can easily be analysed from a range of perspectives, whereas the labor intensiveness of traditional methods is a highly restricting factor. Our approach is possible to create a more accurate description of the knowledge companies actually possess, detecting more reliably latent patterns, such as emerging technologies, whereas human labeling dependent approaches take a rear-view mirror approach to new-to-the-world phenomenon (Suominen and Toivanen, 2015). This has implications for corporate knowledge and patent portfolio management and for example a mergers and acquisitions situation, where a view on the complementarity of patent portfolios can be assessed. A recent example from our sample firms is the acquisition of Alcatel-Lucent by Nokia
So to the question at hand, does Nokia Alcatel-Lucent merger make sense technology wise. Looking at technology, this acquisition might be hard justification. But you are free to evaluate on your own. The Figure above is available from my website, allowing you to look at companies in topics pre an post acquisition.