=Paper=
{{Paper
|id=Vol-1953/healthRecSys17_paper_14
|storemode=property
|title=Hybrid Collaboration Recommendation from Bibliometric Data - the Medical Technology Perspective
|pdfUrl=https://ceur-ws.org/Vol-1953/healthRecSys17_paper_14.pdf
|volume=Vol-1953
|authors=Mark Bukowski,André Calero Valdez,Martina Ziefle,Thomas Schmitz-Rode,Robert Farkas
|dblpUrl=https://dblp.org/rec/conf/recsys/BukowskiVZSF17
}}
==Hybrid Collaboration Recommendation from Bibliometric Data - the Medical Technology Perspective==
https://ceur-ws.org/Vol-1953/healthRecSys17_paper_14.pdf
Hybrid collaboration recommendation from bibliometric data
The medical technology perspective
Mark Bukowski André Calero Valdez Martina Ziefle
AME – RWTH Aachen University HCIC – RWTH Aachen University HCIC – RWTH Aachen University
bukowski@ame.rwth-aachen.de valdez@comm.rwth-aachen.de ziefle@comm.rwth-aachen.de
Thomas Schmitz-Rode Robert Farkas
AME – RWTH Aachen University AME – RWTH Aachen University
smiro@ame.rwth-aachen.de farkas@ame.rwth-aachen.de
ABSTRACT 1 INTRODUCTION
Medical product development is becoming more and more complex Recommender systems (RecSys) in the health domain typically
and requires highly-specialized and interdisciplinary collaborations. address either health professionals or end users (patients). This
Their success relies essentially on the selection of suitable partners. means that the recommended items are typically foodstuffs [5, 14],
However, how to find suitable partners and how to match capabili- sport activities [11], medicine or in some cases even diagnoses. The
ties of an unknown partner with complex project requirements? goals are clear: for example, healthy food predominantly aims at
Suitability must at least be judged with respect to professional burning calories and sport aims at improving physical activity — the
competencies, collaboration capability and project-specific require- recommendations are about behavioristic and behavioral changing
ments — none of which are easily determined. So, partner selection aspects [12]. Common to these types of recommendation is the
is mostly dominated by regional proximity or even coincidence. large field of potential users and their health records [19]. who
This is a typical scenario for recommender systems. Therefore, we contribute sufficient data and thus the knowledge of the RecSys.
aim at discovering the unexploited potential of collaboration part- However, in another health-related field of application, this is
ners by proposing a novel recommendation approach that merges not feasible: recommending collaboration partners in medi-
trust with health-sensitive semantic information. This hybrid ap- cal technology. Here, the amount of active users is limited to
proach should help to identify collaborators matching complex researchers, clinicians and enterprises. Apart from that, the objec-
project requirements faster, better and more holistically. tives are substantially more unclear but also complex due to their
multidimensionality and they need to be selected with regard to
CCS CONCEPTS a project goal and project team (e.g., from physicians, natural and
• Information systems → Decision support systems; Recom- computer scientists to engineers having different professional and
mender systems; Data mining; • Applied computing → Health social capabilities, research habits and objectives).
care information systems; Health informatics; The decisive advantage is that medicine has a semantically struc-
tured terminology (e.g., ICD-10, UMLS). This enables the classifi-
KEYWORDS cation of documents (e.g., scientific publications) with supervised
Health Recommender Systems; Health Informatics; Collaboration learning to extract well-defined feature vectors on which RecSys
Recommendation; Hybrid-Recommendation Interventions; may be based. Therefore, we could perform recommendations with
respect to the technological, product-related and clinical suitabil-
ACM Reference format:
ity of partners. However, it is not sufficient to only rely on this
Mark Bukowski, André Calero Valdez, Martina Ziefle, Thomas Schmitz-
in order to find an appropriate partner — the collaboration capa-
Rode, and Robert Farkas. 2017. Hybrid collaboration recommendation from
bibliometric data. In Proceedings of the Second International Workshop on bilities as a subset of social competences and homophily are of
Health Recommender Systems co-located with ACM RecSys 2017, Como, Italy, importance, too [13]. These can be derived as collaboration trust
August 2017, 3 pages. from the bibliometric meta data: Who worked with whom on which
topic [3]?
Therefore, we propose a hybrid recommender approach that
ties both aspects together: trust-based recommendation [9] based
on collaboration data and semantically structured feature vectors
based on scientific corpora. This should enable the identification
of project-specific, suitable collaboration partners and to recom-
mend them even with fuzzy project goals as support for science
management.
In Proceedings of the Second International Workshop on Health Recommender Systems
co-located with ACM RecSys 2017, Como, Italy, August 2017 (RecSys’17), 3 pages.
© 2017 Copyright for the individual papers remains with the authors. Copying permit-
ted for private and academic purposes. This volume is published and copyrighted by
its editors.
HealthRecSys’17, August 2017, Como, Italy Bukowski et al.
A blood-free scalpel. Medical technology is on the one hand 2.1 Hybrid, Trust-based Group-Recommender
highly complex and diverse, on the other hand it is characterized by After integrating professional competencies, another important
the need for innovation-driven and fast development [17]. The tar- factor must be included. Researchers have a highly unique way of
get group interested in finding project-suitable partners is widely collaborating. Not everyone would/could/should collaborate with
spread: clinic, research, economy. The knowledge transfer is re- each other. We have successfully used bibliometric-based recom-
quired but the identification of suitable partners can be difficult — mendations to identify collaborators in a research cluster, in which
even within the same university. Imagine the following scenario: we used graph mining on the co-authorship graph [15] to deter-
A surgeon contacts the medical technology department with a re- mine interdisciplinary experience (see Fig. 2). Combining such
quest: “During operations too much blood obstructs my view. Can approaches with content-based recommendations [18] should yield
you come up with a scalpel that cuts without bleeding?”1 The med- researchers with topic, method and skill that are complementary
ical demand is clear, but the physician does not have insights into in cross-domain groups [13].
the technological challenges and product development. Medical Social recommendations have been shown to provide higher
technology faces the difficulty to find partners from opaque require- accuracy than mere tag-based approaches [6] and outperform pure
ments: It is not clear what the best approach is and who is able to content-based approaches as they provide additional context to the
implement it. It remains challenging to match project objectives recommendation algorithm. Social network approaches for collabo-
with potential collaboration partners. ration suggestions have already been successfully tested in social
networks for scientists [1] and also in co-authorship networks [8].
2 CAPABILITY MATCHING Still, one further problem remains. Identifying individuals that
One major requirement to meet project goals are collaboration could collaborate is simpler than suggesting collaborators for a
partners with specific professional competences. For this, scien- whole group of researchers. However, the field of group recom-
tific publications, patents and project descriptions are knowledge mendation provides algorithms [4] that consider the trade-off
bases that are directly related to the authors’ in-domain activity between individual and group preferences and can be applied here.
and proficiency [16]. Due to the interdisciplinary characteristics Typical applications are, for example, group recipe recommenda-
of medical technology, the different kinds of researchers leave se- tions [6]. A tensor-based approach seems fruitful in order to com-
mantic tracks from basic research to the application of innovative bine these approaches.
products. The goal is not only to follow these tracks, but also to
process the information and aggregate it to representative feature 2.2 Evaluation
vectors for professional competence recommendation.
Although finding good recommendations is difficult already, con-
The systematized domain language also used in scientific cor-
firming these recommendations is even more difficult. Researchers
pora facilitates semantic text mining: e.g., with the well-established
have very little time to evaluate such systems. Besides other in-
Support Vector Machine. The basis for a clear classification and
formation retrieval methods of evaluation, we are planning to use
comparable representation of professional competences is a domain
intelligible visual representations [10] of our feature vectors. This
model (see Fig. 1).
approach should simplify evaluation in such complex scenarios [7].
The overall aim of this project is to find suitable collaborators
▪ Microsystems ▪ Imaging that contribute complementary skill-sets for a diverse set of require-
technology ▪ Implants and
▪ Nanotechnology prosthetics ments based on collaboration requests including textual descrip-
▪ Biotechnology ▪ Telemedicine
tions and further context-dependent features. Such a system could
Techno- Product-
▪ Photonics logical related ▪ Surgical
▪ Tissue engineering interventions increase the speed of medical technology development and lastly
▪ Materials science ▪ In-vitro diagnostics
▪ Information and ▪ Special therapy and benefit research, patients and society as a whole.
communication diagnostic systems
technology Clinical
ACKNOWLEDGMENTS
▪ Neoplasms This research was supported by Klaus Tschira Stiftung gGmbH.
▪ Mental and behavioral disorders
▪ Diseases of the circulatory system
▪ Diseases of the respiratory system REFERENCES
▪ Diseases of the digestive system [1] Nesserine Benchettara, Rushed Kanawati, and Céline Rouveirol. 2010. A su-
▪ Diseases of the musculoskeletal system
▪ Injury, poisoning
pervised machine learning link prediction approach for academic collaboration
recommendation. In Proceedings of the fourth ACM conference on Recommender
Figure 1: The three-dimensional domain model has 20 systems. ACM, 253–256.
[2] Simon Bruns, André Calero Valdez, Christoph Greven, Martina Ziefle, and Ulrik
technological, product-related and clinical fields classifying Schroeder. 2015. What should I read next? A personalized visual publication
medical technology innovations. recommender system. In International Conference on Human Interface and the
Management of Information. Springer, Cham, 89–100.
[3] André Calero Valdez, Anne Schaar, Martina Ziefle, Andreas Holzinger, Sabina
We trained a generalizing text mining system that classifies Jeschke, and Christian Brecher. 2012. Using mixed node publication network
documents into this domain model with high accuracy (>80%). This graphs for analyzing success in interdisciplinary teams. Active Media Technology
(2012), 606–617.
information needs to be aggregated to a profile-centric feature [4] Yen-Liang Chen, Li-Chen Cheng, and Ching-Nan Chuang. 2008. A group rec-
vector for professional competence as part of the RecSys. ommendation system with consideration of interactions among group members.
Expert systems with applications 34, 3 (2008), 2082–2090.
[5] David Elsweiler, Christoph Trattner, and Morgan Harvey. 2017. Exploiting Food
1 Such a scalpel exists: A laser-scalpel.
Choice Biases for Healthier Recipe Recommendation. (2017).
Hybrid collaboration recommendation from bibliometric data HealthRecSys’17, August 2017, Como, Italy
Figure 2: Using bibliometric data, trust relationships can be inferred from co-authorship. Possible collaborators should have
stronger connections [2].
[6] Georg Groh and Christian Ehmig. 2007. Recommendations in taste related on Knowledge discovery and data mining. ACM, 1285–1293.
domains: collaborative filtering vs. social filtering. In Proceedings of the 2007 [14] Christoph Trattner and David Elsweiler. 2017. Investigating the healthiness
international ACM conference on Supporting group work. ACM, 127–136. of internet-sourced recipes: implications for meal planning and recommender
[7] Anthony Jameson, Martijn C Willemsen, Alexander Felfernig, Marco de Gemmis, systems. In Proceedings of the 26th International Conference on World Wide Web.
Pasquale Lops, Giovanni Semeraro, and Li Chen. 2015. Human Decision Making International World Wide Web Conferences Steering Committee, 489–498.
and Recommender Systems. Recommender Systems Handbook 54 (2015), 611–648. [15] André Calero Valdez, Matthias Dehmer, and Andreas Holzinger. 2016. Applica-
[8] Jing Li, Feng Xia, Wei Wang, Zhen Chen, Nana Yaw Asabere, and Huizhen tion of graph entropy for knowledge discovery and data mining in bibliometric
Jiang. 2014. Acrec: a co-authorship based random walk model for academic data. Mathematical Foundations and Applications of Graph Entropy 6 (2016), 174.
collaboration recommendation. In Proceedings of the 23rd International Conference [16] André Calero Valdez, Denis Özdemir, Mohammed Amin Yazdi, Anne Kathrin
on World Wide Web. ACM, 1209–1214. Schaar, and Martina Ziefle. 2015. Orchestrating collaboration-using visual col-
[9] Paolo Massa and Paolo Avesani. 2007. Trust-aware recommender systems. In laboration suggestion for steering of research clusters. Procedia Manufacturing
Proceedings of the 2007 ACM conference on Recommender systems. ACM, 17–24. 3 (2015), 363–370.
[10] Belgin Mutlu, Eduardo Veas, and Christoph Trattner. 2017. Tags, Titles or Q&As? [17] Stefan Weigel. 2011. Medical Technology’s Source of Innovation. European
Choosing Content Descriptors for Visual Recommender Systems. (2017). Planning Studies 19, 1 (2011), 43–61. DOI:http://dx.doi.org/10.1080/09654313.
[11] Heleen Rutjes, Martijn C Willemsen, and Wijnand A IJsselsteijn. 2016. Un- 2011.530391
derstanding effective coaching on healthy lifestyle by combining theory-and [18] Robert Wetzker, Winfried Umbrath, and Alan Said. 2009. A hybrid approach to
data-driven approaches. In Proceedings of the Personalization in Persuasive Tech- item recommendation in folksonomies. In Proceedings of the WSDM’09 Workshop
nology Workshop, Persuasive Technology 2016. CEUR-WS.org, 116. on Exploiting Semantic Annotations in Information Retrieval. ACM, 25–29.
[12] Hanna Schaefer, Santiago Hors-Fraile, Raghav Pavan Karumur, André Calero [19] Martin Wiesner and Daniel Pfeifer. 2010. Adapting recommender systems to the
Valdez, Alan Said, Helma Torkamaan, Tom Ulmer, and Christoph Trattner. 2017. requirements of personal health record systems. In Proceedings of the 1st ACM
Towards Health (Aware) Recommender Systems. Proc. of DH 17 (2017). International Health Informatics Symposium. ACM, 410–414.
[13] Jie Tang, Sen Wu, Jimeng Sun, and Hang Su. 2012. Cross-domain collaboration
recommendation. In Proceedings of the 18th ACM SIGKDD international conference