=Paper=
{{Paper
|id=Vol-1878/article-01
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-1878/article-01.pdf
|volume=Vol-1878
}}
==None==
https://ceur-ws.org/Vol-1878/article-01.pdf
Towards an Infrastructure for Understanding
and Interlinking Knowledge Co-Creation in
European Research
Diana Maynard1 , Adam Funk1 , and Benedetto Lepori2,3
1
Department of Computer Science, University of Sheffield,
211 Portobello, Sheffield, UK
2
Faculty of Communication Sciences, Universit della Svizzera italiana,
6904 Lugano, Switzerland
3
Laboratoire Interdisciplinaire Sciences, Innovations et Socits (LISIS),
University of Paris Est, 77454 Marne-la-Valle Cedex 02, France
Abstract. This paper describes the initial development of an infras-
tructure for understanding and visualising knowledge co-creation in Eu-
ropean research. Datasets containing information about projects, pub-
lications and patents are enhanced with semantic information enabling
indicators to be generated, that are used to inform users about the state
of art in the research area and new trends. This helps to resolve the prob-
lem of increasing complexity and multidisciplinarity of emerging scien-
tific and technological research. Ontologies and Semantic Web techniques
play a central role in mapping between topics, data, and user queries so
that information can be enhanced, interlinked and aggregated.
Key words: ontologies, linked data, knowledge co-creation, NLP
1 Introduction
Mapping the development of Science and Technology (S&T) has become an issue
of increasing societal and political relevance and, at the same time, of increasing
complexity in terms of measures and indicators. On the one hand, EU policy is
focusing on useful knowledge, in terms of (1) Key Emerging Technologies (KET),
which foster growth and economic development, and (2) knowledge production
to respond to so-called Societal Grand Challenges (SGC), i.e. those challenges
considered critical for the future of European society [1]. On the other hand,
knowledge production no longer comes from one source (i.e. universities), but
emerges from multiple sources. Furthermore, knowledge production processes are
becoming more and more hybrid, crossing disciplines (especially for key technolo-
gies), institutional borders (especially for societal challenges), and geographical
borders (internationalisation).
These developments imply a number of challenges such as integrating a
broader scope of data sources, the construction of more flexible classification
systems, and more comprehensive coverage of actors involved in knowledge co-
creation. Traditional indicators on S&T development fall short of addressing
2
these challenges for various reasons: they tend to be aggregated and focused on
traditional research outputs, like publications and patents, and therefore cannot
adequately identify new places of social innovation and the outreach of research;
they adopt stable and, usually, rather coarse classification schemes (like field of
science or ISI subject categories), which do not allow mapping the emergence of
new knowledge domains at the intersection of existing scientific fields; third, by
the nature of the data used, tend to be backward-looking and thus difficult to
use for policy decisions.
Innovative approaches have been developed in order to overcome some of
these limitations: this includes the use of linguistic techniques and overlay maps
to identify clusters of publications or patents associated with a specific topic [14,
12]; the use of altmetric data to provide a broader view on research impact [17]
and analyse user attention on scientific activities through social media [13]; and
finally, content analysis performed on research project descriptions, as these are
much nearer to the frontier of knowledge production.
However, one of the major problems lies in making connections between the
different kinds of data, and in the provision of cross-cutting search facilities that
enable information from them to be combined. Policy makers often struggle to
get a good picture of the fast-changing nature of research, because information
is dynamic, conflicting and hard to understand. In particular, the language used
can differ widely, and this restricts search usefulness.
The goal of the recently-launched H2020 project KNOWMAK (Knowledge
in the Making in the European Society) is to address these challenges by devel-
oping a web-based tool providing interactive visualisations and state-of-the-art
indicators on knowledge co-creation in the European research area. It will pro-
vide information to users wishing to understand the nature of and connections
between key European research in particular domains (topics), institutions and
locations. The central aspect of this system is a set of ontologies, which form
the bridge between knowledge sources and user queries. These ontologies are
focused on two research areas: Societal Grand Challenges (SGC), and Key En-
abling Technologies (KET). At the heart of this is the quadruple helix model of
innovation, linking government, industry, academia and civil participants.
On querying the KNOWMAK platform, users will be able to identify via the
ontologies some related concepts, including those from the scientific and tech-
nological vocabulary, and to link them with the relevant markers of knowledge
production, specifically publications, patents, research projects and social inno-
vation projects. In this way, they will be also able to visualize maps of space and
actors relevant for that topic, and to zoom in onto specific research projects and
outputs relevant for their topic.
2 Related Work
KNOWMAK builds on work performed in the EU RISIS project4 , which is
creating a distributed infrastructure for research and innovation data and poli-
4
http://risis.eu
3
cies, focusing on three main areas: creating and making available open datasets
on research and innovation issues; developing open platforms for building and
manipulating such datasets; and developing tools to facilitate interconnections
between heterogeneous existing datasets. Critically, the addition of the ontolog-
ical element, which is lacking in RISIS, will enable better integration of all the
different resources, and improved handling of the complexities of language.
While there is a long tradition in using ontologies for addressing policy issues
(see for example [7]), KNOWMAK is faced with a number of specific issues. First,
KET and SGC are themselves rather general terms. and a clear and consistent
delineation from policy documents cannot be expected. Second, KNOWMAK
ontologies need to connect user queries, mostly formulated in rather generic
terms and variable over time and user groups, with the more specialized vo-
cabularies used in scientific documents. To address these issues, KNOWMAK
is developing a staged approach to ontology development, relying first on ex-
isting mapping studies and classifications and on expert knowledge in order to
construct a basic ontology structure and, second, enriching and populating the
ontology vocabulary through text mining techniques on a selection of policy
documents, publications, patents and descriptions of research projects.
The use of text mining techniques for understanding and enhancing inno-
vation has been studied extensively by [4] and [6], who found that traditional
sources of information about R&D activities (reported on company websites and
in databases of patents and publications) could be enhanced significantly by such
methods. However, they only used quite simple methods of analysis, such as reg-
ular expression-based keyword search. In the highly dynamic field of NLP, [2]
used predictive modelling to predict the key technical NLP terms of the future.
Other work has investigated the need for combining information from related
fields to populate domain-specific ontologies, e.g. in the field of metabolomics
[15]. Previous semantic annotation work has demonstrated the power of combin-
ing text mining and ontologies to discover and link information from large-scale
documents such as patent data [16], archived material [10] and social media [11].
Another starting point is the work of [3], who constructed a mapping between
the IPC patent classification codes and 6 of the 7 Societal Grand Challenges.
This mapping forms the basis for part of our ontology. Furthermore, they devel-
oped a set of subcategories for each SGC, through a comprehensive expert-based
process, including feedback and validation from the European Commission. How-
ever, this creates new problems, since the mapping is still evolving, and the sub-
classes generated for the SGCs do not necessarily match our needs. Our ontology
needs to relate to the language and needs of the users, the EU policies, and the
projects, patents and publications data. It also needs to be easily extendable as
new kinds of information sources are added: later in the project, information col-
lected from social media will be added, which will almost certainly be expressed
in different kinds of language. Furthermore, beyond the life of the project, the
kinds of information covered could extend both beyond European research to
international research, and beyond the SGC and KET areas to other fields of
research.
4
Fig. 1. KNOWMAK platform architecture
3 The KNOWMAK platform
Figure 1 shows the basic architecture of the framework, connecting the user
queries (at the bottom of the diagram) to the information in the databases (at
the top of the diagram) via a set of enhanced databases and indicators (in the
middle) which allow users to explore the data. The enhanced databases connect
relevant information from the projects, patents, publications and policies to the
ontologies, which in turn allows indicators to be created that can provide answers
to users’ questions and a means to visualise associated data.
One of the major problems with linking the data sources to the ontologies
is that the language is very different in each of these kinds of data. This means
that finding relevant instances of the ontology classes in the data (semantic
annotation) is tricky. Standard NLP techniques for ontology population operate
on principles such as term recognition [8], but terms in these documents may be
completely different and may also change over time. We will discuss this further
in Section 4. Another problem is that the information in the original databases
is not always standardised or correct. For example, the same name or term is
written differently, mistakes occur and so on. In both patent and publication
5
databases, this is a long-standing issue, and although attempts have been made
to address this through (a) database cleaning and enhancement and (b) use of
NLP and ML techniques to match and link names and terms, it is far from
resolved.
4 Ontology development
The ontology acts as a bridge between the users and the data stored in our knowl-
edge bases, enabling dynamic linking between them. Documents are annotated
with one or more classes from the ontology, linking them with topics which are
used, along with geographical and actor information (at which location was the
research done, and by which institution(s)) to build indicators about the data.
The ontology is designed via an expert process, taking into consideration the
needs of the users (following dedicated workshops) and previous existing classifi-
cations and protocols, following a principled development process. The ontology
may be constantly refined depending on new information: the databases get up-
dated, terms are dynamic, and user needs change. A first version of the ontology
is available online5 , although it is in continuous update.
From a methodological perspective, two main issues need to be addressed.
First, the structure of the ontologies is problematic. Users may search for quite
general topics like “Societal Grand Challenges” which need to be associated with
a more fine-grained vocabulary to allow for the identification of relevant data.
Since the two areas of KET and SGC are rather different, they are developed
independently, though mappings are created between relevant classes. Because
the ontologies form the central hub of the system, they must cater to the needs
of both the users (i.e. to cover the topics that they might want to search for) and
the research domain. Through the use of sophisticated NLP and term recognition
technologies linking the sources to the ontologies, we aim to help mitigate the
problem of different terminologies. In the past, only use of keywords and simple
matching algorithms to the text has typically been used. More complex document
similarity, domain modelling and linguistic techniques will therefore be used to
complement these strategies.
The second issue is the annotation of the data with the ontologies (seman-
tic annotation). The difference between the language of the policies and data is
again critical. Related terms are identified using methods for capturing lexical
ambiguity and variation [11], and for the identification of related terms (e.g.
semantic frames, word embeddings and ontology generation techniques [9]). Ad-
ditionally, the documents retrieved by applying ontologies will be used as seeds
in order to extend the search, either by textual similarity or by exploiting the in-
ternal structure of the data, for example with the use of overlay maps to identify
clusters of publications or patents associated with a specific topic [14].
We take as a starting point some existing classifications, which we merge
and map. For example, there exist already mappings between IPC (Interna-
tional Patent Classification) codes and both KETs and SGCs [18, 3]. This has
5
https://gate.ac.uk/projects/knowmak/
6
the advantage that when new patents are added to the database, they can be
automatically connected to the ontology. However, as [18] discovered, these are
not sufficient to cover all cases, so an annotation process is still necessary to
supplement this.
For the KET ontology, we also make use of the structure implemented in
the nature.com ontologies portal [5]. This provides a repository for the semantic
schemas driving the Nature.com publishing platform and datasets, comprising
a common network of inter-related and constantly evolving ontologies. Three of
the KETs are represented here: nanotechnologies, industrial biotechnology and
photonics. Some subjects are also connected to DBpedia and MESH. Linking
with the nature.com ontology helps with mapping publications, and enables
future extension of our ontology to other topics. For the other 3 KETs, we
manually create subclasses based on the EU policy documents which describe
how the KETs are structured.
One major issue is that the topics covered in the 6 KETs overlap considerably.
Aside from the cross-cutting KET AMT (Advanced Manufacturing Techniques),
there are also many overlaps between topics in Photonics (e.g. miniaturisation
issues with modules for fibre-optics), Advanced Materials (miniaturisation of
various AM for Nano- and Micro-Electronics); and Advanced Manufacturing
Techniques (e.g. semiconductor manufacturing processes). These are modelled
in our ontology through multiple inheritance (e.g. AMT Advanced Materials is
a subclass of both AMT and Advanced Materials).
For the projects data, we make use of the existing subject index6 developed in
the RISIS project to map some documents to our ontology (for example, “mate-
rials technology” can be mapped to the Advanced Materials KET, while “nan-
otechnology and nanoscience” can be mapped to the Nanotechnologies KET.
Again, for the rest, we need to use text mining tools to annotate documents
against the relevant ontology class.
The EU H2020 policy documents divide each of the 7 SGCs into sub-categories.
For example, bioeconomy has the 4 subcategories: food security, blue growth,
agricultural research and innovation, and bio-economy. Figure 2 shows a portion
of the SGC ontology, including some of these mappings. For example, “A61K
39” is the patent code referring to “Medicinal preparations containing antigens
or antibodies”. We can also see in the ontology that “Biotech” is a subclass of
both the Health and Bioeconomy SGC top-level classes.
5 Challenges and Future Directions
The role of the ontology in the KNOWMAK framework is critical, yet it is
also hard to evaluate. While many ontology evaluation methodologies have been
proposed, which validate the rigour, correctness and coverage of the ontology,
by far the most useful method is a task-based evaluation. In other words, does
the ontology fulfil the needs of the users? In terms of rigour and correctness,
6
https://ec.europa.eu/growth/tools-databases/kets-tools/sites/default/
files/library/final\_report\_kets\_observatory\_en.pdf
7
Fig. 2. Part of the SGC ontology
relying on existing established ontologies such as nature.com and the various
mapping efforts goes some way to validating this. The construction process also
relies heavily on continuous feedback from real users and knowledge experts: the
ontology must cover the kinds of questions they want to ask, and must return
useful results. This will be monitored throughout the course of the project, but
it remains a risk.
In summary, the KNOWMAK project, while founded on some robust princi-
ples and rooted in some established previous initiatives, nevertheless attempts to
address some very real and challenging problems. Indeed, the need for this type
of work is precisely because there are so many issues in opening and linking dif-
ferent kinds of European research activity, and because of the dynamic nature of
these emerging technologies. The framework is currently in early stages of devel-
opment, and there remains much still to be done, so this paper serves primarily
to set out the key motivations, general architecture design, and challenges to
be overcome, and to position the work within a wider context. It is clear that
the Semantic Web community can offer much in the way of existing solutions to
the problems of ontology linking and population, which can help with the prob-
lem of bridging unstructured and structured content using semantic annotation
techniques, and which will form part of our future work.
Acknowledgements This work was partially supported by the European
Union under grant agreement No. 726992 KNOWMAK.
References
1. (EC), E.C.: Europe 2020: A strategy for smart, sustainable and inclusive growth.
Working paper {COM (2010) 2020} (2010)
2. Francopoulo, G., Mariani, J., Paroubek, P., Vernier, F.: Providing and Analyzing
NLP Terms for our Community. Computerm 2016 p. 94 (2016)
8
3. Frietsch, R., Neuhäusler, P., Rothengatter, O., Jonkers, K.: Societal grand chal-
lenges from a technological perspective: Methods and identification of classes of
the international patent classification ipc. Tech. rep., Fraunhofer ISI Discussion
Papers Innovation Systems and Policy Analysis (2016)
4. Gök, A., Waterworth, A., Shapira, P.: Use of web mining in studying innovation.
Scientometrics 102(1), 653–671 (2015)
5. Hammond, T., Pasin, M.: The nature.com ontologies portal. In: 5th Workshop on
Linked Science (2015)
6. Kahane, B., Mogoutov, A., Cointet, J.P., Villard, L., Laredo, P.: A dynamic query
to delineate emergent science and technology: the case of nano science and tech-
nology. Content and technical structure of the Nano S&T Dynamics Infrastructure
pp. 47–70 (2015)
7. Loukis, E.N.: An ontology for g2g collaboration in public policy making, imple-
mentation and evaluation. Artificial Intelligence and Law 15(1), 19–48 (2007)
8. Maynard, D., Funk, A., Peters, W.: SPRAT: a tool for automatic semantic pattern-
based ontology population. In: International Conference for Digital Libraries and
the Semantic Web. Trento, Italy (September 2009)
9. Maynard, D., Funk, A., Peters, W.: Using Lexico-Syntactic Ontology Design Pat-
terns for ontology creation and population. In: WOP 2009 – ISWC Workshop on
Ontology Patterns. Washington, USA (October 2009)
10. Maynard, D., Greenwood, M.A.: Large Scale Semantic Annotation, Indexing and
Search at The National Archives. In: Proceedings of LREC 2012. Turkey (2012)
11. Maynard, D., Greenwood, M.A., Roberts, I., Windsor, G., Bontcheva, K.: Real-
time social media analytics through semantic annotation and linked open data. In:
Proceedings of WebSci. Oxford, UK (2015)
12. Rafols, I., Porter, A.L., Leydesdorff, L.: Science overlay maps: A new tool for
research policy and library management. Journal of the American Society for in-
formation Science and Technology 61(9), 1871–1887 (2010)
13. Robinson-Garcı́a, N., Torres-Salinas, D., Zahedi, Z., Costas, R.: New data, new
possibilities: exploring the insides of altmetric.com. arXiv preprint arXiv:1408.0135
(2014)
14. Rotolo, D., Rafols, I., Hopkins, M.M., Leydesdorff, L.: Scientometric mapping as
a strategic intelligence tool for the governance of emerging technologies. (2014)
15. Spasić, I., Schober, D., Sansone, S.A., Rebholz-Schuhmann, D., Kell, D.B., Paton,
N.W.: Facilitating the development of controlled vocabularies for metabolomics
technologies with text mining. BMC bioinformatics 9(5), S5 (2008)
16. Tablan, V., Bontcheva, K., Roberts, I., Cunningham, H.: Mı́mir: an open-source
semantic search framework for interactive information seeking and discovery. Jour-
nal of Web Semantics 30, 52–68 (2015), http://dx.doi.org/10.1016/j.websem.
2014.10.002
17. Thelwall, M., Haustein, S., Larivière, V., Sugimoto, C.R.: Do altmetrics work?
twitter and ten other social web services. PloS one 8(5), e64841 (2013)
18. Van de Velde, E.: Feasibility study for an EU monitoring mechanism on key en-
abling technologies. IDEA Consult (2012)