analysis allows decision makers to investigate the experiences of other (possible competitor)
institutions and/or countries, in such a way as to know the past and the current R&D activities in
the fields of interest, to delineate their evolution and to foresee their future developments.
Furthermore, patent analysis allows the construction of a detailed picture of the R&D
cooperations among different institutions and/or countries and can be an indicator of geo-
political evolutions happening all over the world (Bornmann and Daniel, 2008; Cooke and Hall,
2013; Shen et al., 2017).
Most of the past approaches for patent analysis were based on classical statistics.
However, the impressive development of innovations in all the R&D fields is leading to a
huge increase of patent data. Therefore, it is reasonable to foresee that, in the next future,
Big Data-centered techniques will be compulsory to fully exploit the potential of patent data.
In this last scenario, the adoption of approaches based on network analysis is extremely
promising (Cammarano et al., 2015; Clauset et al., 2004, 2009; Leicht et al., 2006; Wasserman
and Faust, 1994; Zardi et al., 2016). As a matter of facts, network analysis allows a full
comprehension and a complete management of those phenomena where relationships
among objects to investigate play the key role and, at the same time, the corresponding
variables are strictly related to each other. This is exactly the future scenario characterizing
patent and innovation management, and, at the same time, it is the “worst-case scenario”for
classic statistic-based approaches, which present several limitations when operating therein
(Tsvetovat and Kouznetsov, 2011).
As a confirmation of the adequacy of network analysis for patent investigation, in
the past literature, several approaches to facing this issue can be found (see, for instance,
Cammarano et al., 2015; Ellis et al., 1978; Hu and Zhang, 2017; Hung and Wang, 2010;
Yang et al., 2015).
Centrality is one of the most investigated issues in network analysis. It aims at
measuring the importance of a node in a network. It allows experts: to measure the relevance
and the criticity of nodes in their networks; to define forms of distance between network
nodes or areas; to measure the cohesion degree of a sub-network; and to identify cohesive
sub-networks or network communities.
In the past, several centrality measures have been proposed in the literature (Brin and
Page, 1998; Chen et al., 2013; Freeman, 1977, 1979; Hage and Harary, 1995; Sabidussi, 1966;
Stephenson and Zelen, 1989). Among them, the most general and best known ones are:
degree centrality, based on the number of arcs incoming in, or outgoing from, each node;
closeness centrality, based on distances between nodes; betweenness centrality, based on
the shortest paths connecting pairs of nodes; and eigenvector centrality, based on both the
number and the centrality of nodes whose outgoing arcs are incident on the nodes of
interest. All these measures, as well as the other ones proposed in the literature, could be
adopted in the investigation of patents. However, they are not tailored to this scenario and
could return approximate results. This is because patents have a very relevant peculiarity
that is notfound elsewhere ( for instance, inscientific papers, Ferrara et al., 2018), in that, if a
cites a patent p
, then p
looses a part of its value.
If we report this reasoning to the network analysis context, we have that, for a node,
having incoming arcs is extremely positive, whereas having outgoing arcs is negative. Past
centrality measures certainly distinguish between these two kinds of arc; for instance,
degree centrality distinguishes between indegree and outdegree (Hanneman and Riddle,
2005). However, they do not combine centrality values originated from the incoming arcs
with those derived from the outgoing ones. What is missing is precisely a centrality measure
that first assigns a positive ranking to incoming arcs and a negative ranking to outgoing
ones and, then, combines these rankings to obtain a unique value.
In this paper, we aim at providing a contribution in this setting. In fact, we propose a
well-tailored centrality measure for evaluating patents and their citations.