Retrieving Specialisation Profiles for Patent Service Providers – A Use Case of the Relative Specialisation Index

Poster (PDF Available) · September 2016with 270 Reads
DOI: 10.13140/RG.2.2.19293.82408
, Jena, 2016 Annual Assembly of the Michael Stifel Centre for Data-Driven and Simulation Science Jena (MSCJ), DOI:10.13140/RG.2.2.19293.82408
Cite this publication
Abstract
Specialisation profiles are an essential tool in matchmaking between professional service providers (agents) and potential clients. In a heterogenous market such as patent services, specialisation profiles reduce the cost of information for clients in need of highly specific expertise by reducing the search space to relevant agents. By using the RSI Specialisation we demonstrate how idiotypic and dichotomous specialisation profiles can be retrieved for patent agent firms.
Retrieving Specialisation Profiles for Patent
Service Providers – A Use Case of the
Relative Specialisation Index
Lutz Maicher and Kazimir Menzel
Technology Transfer Research Group
Ernst-Abbe-Platz 2, 07743 Jena, Germany
{lutz.maicher,kazimir.menzel}@uni-jena.de
http://www.tt.uni-jena.de
Why do we need specialisation profiles?
Specialisation profiles are an essential tool in matchmaking between
professional service providers (agents) and potential clients. In a
heterogenous market such as patent services, specialisation profiles
reduce the cost of information for clients in need of highly specific
expertise by reducing the search space to relevant agents.
Defining specialisation profiles
We define an activity profile of an entity as a rank-ordered list of
activities [PTM13]. A collection of such profiles has a matrix structure,
whose entries contain the counts of realisations of an activity by an
entity. Market structure and size of the agents are reconstructed from
the sums over the agent and activity vectors in the matrix. We then
reduce this matrix to a logical matrix where each entry is 1, if an entity
specialises in a given field, and 0 otherwise. We then define the
specialisation profile as the set of activities of an entity, where it is
more active to both its own and its competitors’ average.
The challenge – idiotypic and dichotomous profiles
To satisfy the need for simplicity, a specialisation profile has to be
dichotomous, distinguishing only between specialised and not
specialised activities of an entity. To fulfil the need for meaningful
information, we require a specialisation profile to be idiotypic, i.e.
taking into account further properties of an entity such as size, time in
market or human resources. Data on the latter is unreliable, wildly
scattered and of high variance. We can, however, approximate these by
using our knowledge of the market structure given in the matrix.
Relative vs. Absolute Specialisation
Specialisation in an economic sense is defined as prioritising one field of
activity over one or many others. The literature distinguishes between
two concepts of specialisation, absolute specialisation that compares
counts and relative specialisation that considers market structure and
size [BG04, 741-2]. Obviously, a na¨ıve approach of counting the
observations yields in low-information specialisation profiles. Hence we
turn to the concept of relative specialisation which has its roots in the
theory of the comparative advantage from trade theory [HO06, 677-678].
RSI Specialisation
To build this new classifier, we combine the idiotypic measure ARCA
[HO06, 684], referred to as Relative Specialisation Index (RSI), with its
standard deviation as a threshold to distinguish specialised activities.
(1) We first compute the RSI:
RSIa,c =Aa,c
Pn
c=1 Aa,c
Pm
b=1,b6=aAb,c
Pm
b=1,b6=aPn
c=1 Ab,c
where aand brefer to entities, cto an activity and Ato the matrix
containing the counts of realised activities.
The RSI ranges from [-1,1] with mean 0, where 0 indicates market
average, positive values indicate more and negative less realisations than
the average competitor weighted by the entity’s activity profile.
(2) We then exploit the regular pattern of the positive branch of the RSI
distribution and define the entity specific threshold as its standard
deviation. The entity is specialised in an activity, if the activity’s RSI
score is more than one standard deviation above the mean:
RSI specialisation =(specialised RSIa,c > σ(RSIa)
not specialised RSIa,c σ(RSIa)
Sample and Data
Our data set contains all patent applications that have been published
by the EPO in 2014 and 2015. The applications are assigned to the
patent firms (1,822 in total) (our entities) and to 130 different two-digit
IPC classes (our activities). To avoid distortions, firms are required to
have at least three assignments in a given class as well as at least ten
assignments overall leading to 973 firms with 284,000 assignments.
Examples of RSI-Specialisation Profiles for Patent Agents
0
10000
20000
30000
0 5 10 15
patent assignments to agent per class
patent assignments per class
RSI specialisation
not specialised
specialised
Head Class
A
B
C
D
E
F
G
H
ENF threshold
separating hyperplane
studio−karaghiosoff−e−frizzi−srl
0
10000
20000
30000
0 50 100 150
patent assignments to agent per class
patent assignments per class
RSI specialisation
not specialised
specialised
Head Class
A
B
C
D
E
F
G
H
ENF threshold
separating hyperplane
cbdl−patentanwalte
Studio Karaghiosoff e Frizzi represents a midsized firm, which typically
pursues a niche strategy. We can see that they specialise in seven out of
eight activities, which agrees with our expectations.
CBDL is a highly active firm that typically aims for diversification. Its
profile shows how the RSI distinguishes specialisations from generally
frequent activities as is highlighted by the separating hyperplane.
Application
The IP Industry
Base (http://s.fhg.de/IPIB)
is an analytical database
about patent service
providers. Patent data from
the European Patent Office
(EPO) is continuously
integrated into the database
and analysed. As shown
here, the RSI specialisation
is embedded in the IPIB.
We intend to weekly update
the profiles based on the
patent applications published in the time slice of the last five years.
Further Research
In further research we will investigate the changes of RSI specialisation
profiles over time. The temporal robustness of the RSI specialisation is
important for the quality of the signal created by changes in the
specialisation profiles.
References
Brusoni, Stefano ; Geuna, Aldo:
Specialisation and Integration.
In: Moed, Henk (Hrsg.) ; Gl¨
anzel, Wolfgang (Hrsg.) ; Schmoch, Ulrich (Hrsg.): Handbook of Quantitative Science
and Technology Research.
Springer, 2004, Kapitel 33, S. 733–758
Hoen, Alex R. ; Oosterhaven, Jan:
On the Measurement of Comparative Advantage.
In: The Annals of Regional Science 40 (2006), Nr. 3, S. 677–691
Prilop, Michael ; Tonisson, Liina ; Maicher, Lutz:
Designing Analytical Approaches for Interactive Competitive Intelligence.
In: International Journal of Service Science, Management, Engineering, and Technology 4 (2013), Nr. 2, S. 34–45
This research hasn't been cited in any other publications.
    • Springer
    Springer, 2004, Kapitel 33, S. 733-758
  • Article
    Full-text available
    This article shows that the distribution of the standard measure of revealed comparative advantage (RCA), which runs from 0 to 8, has problematic properties. Due to its multiplicative specification, it has a moving mean without a useful interpretation, while its distribution depends on the number of countries and industries. This article proposes an alternative, additive RCA, running from –1 to +1, with a bell-shaped distribution that centres on a mean equal to zero, independent of the classifications used. Statistical tests show the additive index to be more stable empirically too. Furthermore, the article proposes an aggregate RCA that runs from 0, when pure intra-industry trade prevails, to 1 in the case of pure inter-industry trade. Comparable conclusions hold for the location quotient (LQ), which is used as a measure for the revealed locational attractiveness of certain regions or countries for certain types of industry.