=Paper=
{{Paper
|id=Vol-2030/HAICTA_2017_paper39
|storemode=property
|title=Towards a Global Record of Stocks and Fisheries
|pdfUrl=https://ceur-ws.org/Vol-2030/HAICTA_2017_paper39.pdf
|volume=Vol-2030
|authors=Yannis Tzitzikas,Yannis Marketakis,Nikos Minadakis,Michalis Mountantonakis,Leonardo Candela,Francesco Mangiacrapa,Pasquale Pagano,Costantino Perciante,Donatella Castelli,Marc Taconet,Aureliano Gentile,Giulia Gorelli
|dblpUrl=https://dblp.org/rec/conf/haicta/TzitzikasMMMCMP17
}}
==Towards a Global Record of Stocks and Fisheries==
https://ceur-ws.org/Vol-2030/HAICTA_2017_paper39.pdf
Towards a Global Record of Stocks and Fisheries
Yannis Tzitzikas1,2, Yannis Marketakis1, Nikos Minadakis1, Michalis
Mountantonakis1,2, Leonardo Candela3, Francesco Mangiacrapa3, Pasquale Pagano3,
Costantino Perciante3, Donatella Castelli3, Marc Taconet4, Aureliano Gentile4, Giulia
Gorelli4
1
Institute of Computer Science, FORTH-ICS, Heraklion, Greece, e-mail: {tzitzik, marketak,
minadakn, mountant}@ics.forth.gr
2
Computer Science Department, University of Crete, Heraklion, Greece
3
Consiglio Nazionale delle Ricerche, Pisa, Italy, e-mail: {leonardo.candela,
francesco.mangiacrapa, pasquale.pagano, costantino.perciante, donatella.castelli}@isti.cnr.it
4
Food and Agriculture Organization of the United Nations, Rome Italy, e-mail: {marc.taconet,
aureliano.gentile, giulia.gorelli}@fao.org
Abstract. The collation of information for the monitoring of fish stocks and
fisheries is a difficult and time-consuming task, as the information is scattered
across different databases and is modelled using different formats and
semantics. Our purpose is to offer a unified view of the existing stocks and
fisheries information harvested from three different database sources (FIRMS,
RAM and FishSource), by relying on innovative data integration and
manipulation facilities. In this paper, we describe the activities carried out to
realize the Global Record of Stocks and Fisheries (GRSF) which aims at
offering an integrated and enriched view on data about fish stocks and fisheries
from the database sources. More specifically we describe the model, the
workflow and the software components for producing GRSF records and make
them easily available to the users.
Keywords: fish stock, fishery, semantic data integration, data publication
1 Introduction
Fish Stocks are groups of individuals of a species occupying a well-defined spatial
range independent of other stocks of the same species, e.g. swordfish in the
Mediterranean Sea1. A Fishery is a unit determined by an authority or other entity
that is engaged in raising and/or harvesting fish. Typically, the unit is defined in
terms of some or all of the following: people involved, species or type of fish, area of
water or seabed, method of fishing, class of boats and purpose of activity, e.g.
Fishery for Atlantic cod in the area of East and South Greenland2. Information about
Fish Stocks and Fisheries is widely used for the monitoring of their status, and to
1
http://firms.fao.org/firms/resource/10025/en
2
https://www.fishsource.org/stock_page/688
328
identify appropriate management actions (Hilborn & Walters, 2013), with the
ultimate goal of sustainable exploitation of marine resources. For these reasons
completeness, adequacy and validity of information is crucial. Although this key
role, there is no “one stop shop” for accessing stocks and fisheries data. Such
information is usually collected (and produced as a result of data analysis) by the
fishery management authorities at regional, national and local level. Therefore, the
overall information is scattered across several databases, with no standard structure
due to the specific local needs of the different bodies. Furthermore, the guidelines for
populating existing registries are therefore heterogeneous, and every registry is
actually a “database silo” that is not expected to interoperate with others to offer a
global view on existing information.
Our objective is to construct a Global Record of Stocks and Fisheries (for short
GRSF) capable of containing the corresponding information categorized into
uniquely and globally identifiable records. Instead of creating yet another registry,
we focus on producing GRSF records by using existing data. This approach does not
invalidate the process being followed so far, in the sense that the organizations that
maintain the original data are expected to continue to play their key role in collecting
and exposing them. In fact, GRSF does not generate new data, rather it collates
information coming from the different database sources, facilitating the discovery of
inventoried stocks and fisheries arranged into distinct domains.
The advantages of this approach include: (a) offering increased data coverage
compared to the single sources of information, (b) integrating information and unique
identification of stocks and fisheries coming from the different database sources, and
(c) answering queries that would be impossible to be answered from the individual
database sources. These characteristics meet the needs of the main business cases that
are: (i) supporting the compilation of stock status summaries at regional and global
level and (ii) providing services for the traceability of sea-food products.
In this paper we introduce the process that has been used for constructing and
easily maintaining GRSF. In fact, GRSF maintenance is an almost continuous
activity since data providers can constantly offer new or revised information. The rest
of the paper is organized as follows: Section 2 discusses the motivation and the
requirements. Section 3 describes the architecture and the technical components for
realizing GRSF. Section 4 discusses the current results. Finally, Section 5 concludes
and identifies directions for future work and research.
2 Motivation and Settings
The objective of GRSF is to act as a “one stop shop” for stocks and fisheries
records. It realizes an innovative environment supporting the collaborative
production and maintenance of a comprehensive and transparent global reference set
of stocks and fisheries records, that will boost regional and global stocks and
fisheries status and trend monitoring, as well as, responsible consumer practices. To
this end, a selected set of data sources is exploited for delivering relevant
information. To ensure a high quality final product, a set of guidelines and standards
has been identified which are described later in this section.
329
2.1 The Data Sources
Below we describe the three database sources that have been used so far to
harvest stocks and fisheries information. These sources are (a) Fisheries and
Resources Monitoring System (FIRMS), (b) RAM Legacy Stock Assessment
database, and (c) FishSource. The rationale for the selection of these sources, is that
they contain complementary information (both conceptually and geographically).
More specifically FIRMS is mostly reporting at regional level, while RAM is
reporting at national or subnational level, and FishSource is more focused on the
fishing activities. All of them contribute to the overall aim to build a comprehensive
and transparent global reference set of stocks and fisheries records that will boost
regional and global stocks and fisheries status and trend monitoring as well as
responsible consumer practices. Since the construction of GRSF is an iterative
process, we will support integrating contents from these three sources in early
releases of GRSF, and in future we will investigate exploiting new ones (i.e. FAO
Global Capture Production Statistics database3).
FIRMS (FIsheries and Resources Monitoring System)4 provides access to a wide
range of high-quality information on the global monitoring and management of
stocks and fisheries. It collects data from 14 intergovernmental organizations (that
are partners of FIRMS) and contains information about the status of more than 600
stocks and 300 fisheries. The information provided by the organizations is ingested
in a database and published in the form of XML backboned fact sheets.
RAM (RAM Legacy Stock Assessment Database)5 provides information
exclusively on the fish stocks domain. It is a compilation of stock assessment results
and time series of stock status indicators for commercially exploited marine
populations from around the world. The assessments are assembled from 21 national
and international management agencies for approximately one thousand stocks.
RAM contents are stored in a relational database and are publicly available by
releasing versions of the database in MS Access and Excel format.
FishSource6 compiles and summarizes publicly available scientific and technical
information about the status of fish stocks and fisheries. It includes information about
the health of stocks, the quality of their management, and the impact of fisheries on
the rest of the ecosystem. It is mainly exploited from seafood industry for assisting in
taking the appropriate actions for improving the sustainability of the purchased
seafood. Information in FishSource is organized into fishery profiles associated with
the exploited stocks, and currently contains more than 2,000 fishery profiles.
2.2 Requirements
The selected database sources were originally constructed to fulfil different
requirements and needs. Furthermore, they have been developed and are maintained
from different initiatives. As a result, they are using different standards, data models,
3
http://www.fao.org/fishery/statistics/global-capture-production/en
4
http://firms.fao.org/firms/en
5
http://ramlegacy.org
6
http://www.fishsource.com/
330
conceptualizations and terminologies for capturing similar information. As an
example consider the fish species that are included in a particular stock or fishery;
they can be identified either using (a) their scientific name (e.g. Thunnus albacares),
(b) their common name in any language (e.g. Yellowfin tuna in English), or (c)
standard codes for identifying them (e.g. YFT7). Furthermore, the different data
sources use diverse criteria for identifying the uniqueness of a stock or fishery, as
well as diverse conventions for naming their records.
GRSF aims at harmonizing the harvested information by adopting a set of
standards that have been discussed and agreed with representatives of the database
sources. In particular, these standards have been identified by two technical working
group meetings that have been organized. The working groups have defined which
are the international standards that will be used (e.g. FAO 3Alpha codes for species,
ISO3 country codes for flag states), which values define the uniqueness of a stock or
a fishery record, which values are mandatory to accept a record as a complete one, as
well as guidelines for generating unique and global identifiers (both human and
machine interpretable) and names for the GRSF records. A detailed description of a
GRSF record with respect to those guidelines can be found in Section 2.3.
The main challenge for the construction of the GRSF is the ability to semantically
integrate data coming from different data sources. To tackle this challenge, we
decided to rely on semantic web technologies and use top level ontologies. The best
candidate is the MarineTLO (Tzitzikas et al., 2016-a) which provides (a) consistent
abstractions or specifications of concepts included in all data models or ontologies of
marine data sources and (b) the necessary properties to make GRSF a coherent
source of facts relating observational data with the respective spatiotemporal context
and categorical domain knowledge. The rationale is that we map attributes from
different data sources into classes and properties of the top-level ontologies. To this
end we could also mention works like (Pham et al., 2016) that automate the mapping
process using machine-learning techniques.
2.3 The GRSF record
Each GRSF record is composed of several fields to accommodate the incoming
information and data. The fields can be functionally divided into time-independent
and time-dependent. The first group contains the identification, descriptive and other
information which describes various aspects of a stock or fishery, and the latter
contains the stocks and fishery indicators. In general, there are two types of GRSF
records: (a) stocks and (b) fishery GRSF records. Both types of records share some
common metadata like their time-independent information. Furthermore, records are
assigned information about areas and their original sources. Finally, each record is
assigned several time-dependent information modeled as dimensions. In the case of
stock GRSF records, the dimensions refer to abundance levels and exploitation rates.
In the cases of fishery GRSF records, the dimensions refer to catches and landings
indicators. We could say that a GRSF record resembles a data item in a database and
as such we are describing its corresponding details in the schema shown in Fig. 1.
7
According to FAO 3Alpha code http://www.fao.org/fishery/collection/asfis/en
331
Fig. 1 The STAR schema of a GRSF record
2.4 The Process
The process for constructing GRSF consists of a sequence of steps which are
shown in Fig. 2. Below we describe these steps in detail. The technical components
that carry out each step of the process are described in detail in Section 3.
Fig. 2 The steps required for constructing and exploiting GRSF
Fetch. GRSF does not affect the data from the remote database sources. This
means that the maintainers of the database sources will continue to update them in
their own systems. For including the providers’ data in the GRSF it is important to
periodically fetch the raw data (in their original form) or the data in a different format
or view if they are exposed using particular services (i.e. in other formats like JSON
332
or XML). In particular FIRMS offers a set of services that exposes their contents in
XML format, RAM publish their MS Access database in their website, and
FishSource exposes specific parts of their relational database as JSON data through a
set of web services.
Transform. After fetching the data it is important to transform them so that they
have a similar structure and semantics. At this stage data is transformed from XML,
JSON and MS Access to RDF format. Specifically, data is transformed into instances
of the MarineTLO ontology with respect to the identified GRSF requirements.
Information harvested from the database sources will be mapped to the agreed GRSF
standards, when not already compliant. Furthermore, during this step a set of
proximity rules are applied (using the species, area and gear fields) for identifying
similar records. This creates groupings of similar records that are being used in
subsequent phases (during the curation & validation phase).
Dissect. This step is important for complying with the standards, for traceability
aspects. In some cases, sources contain aggregated information in their records. For
example, in a single fishery record there could be included more than one species,
fishing gears or flag states. These aggregated records are therefore dissected to
produce new GRSF records, each containing one single value for the above-
mentioned fields, and thus complying with the requirements for traceability.
Merge. This step ensures that the contents that have been added in the GRSF
staging database are properly connected based on a set of criteria. This is achieved by
linking records that have the same values on particular fields (specifically time-
independent values) for producing a new single GRSF record. For example, if there
are stock records having the same species and water area, we can merge them into a
single stock. During this process, we also use external knowledge to detect
similarities among different names and terminologies used in the database sources
(i.e. species names). The time-dependent information for the merged records will be
kept distinct although collated and associated to the final merged GRSF record, with
clear indication of the database source and the reference year.
Publish (for curation). The contents of the GRSF staging database are being
replicated into a public GRSF database, which is actually a triple-store. The triple-
store can be used as a reference endpoint for answering complex queries about stocks
and fisheries records. Furthermore the contents are published in a data catalogue
offered through the D4Science (Candela et al., 2014) infrastructure. These resources
allow the experts inspecting the contents of the GRSF and curate them appropriately.
During this step, Universally Unique Identifiers (UUID) and human readable
semantic identifiers are generated and associated to each GRSF record. The former
are generated based on a standard algorithm and are used to uniquely identify
records. The latter are generated using various GRSF fields and populated with
standard codes and allow the identification and interpretation of records by humans.
Curate & Validate. During this step, a community of experts browse over the
GRSF records and curate them in various ways. At this stage, the GRSF records are
in a pending status waiting for approval by a human expert. During this process, the
experts are able to either approve or reject a record, as well as to suggest alternative
processes for merging records and to attach annotations with a narrative text.
333
Publish (for exploitation). The GRSF records that has been approved during the
previous phase are being published into public and read-only databases as final
GRSF products that can be exploited from the communities of interest.
3 Software Components and Architecture
The D4Science infrastructure and gCube technology (Assante et al., 2016) enable
the development of Virtual Research Environments (VREs) that provide the users
with a web-based set of facilities to accomplish various tasks. For the purpose of
GRSF, we developed the appropriate VREs acting as a gateway for the “one stop
shop” for stocks and fisheries records. More specifically we exploit the data
cataloguing facilities of the infrastructure for manipulating and exposing GRSF
records to the wide audience.
The core component for constructing GRSF is MatWare (Tzitzikas et al., 2014).
MatWare is a framework that automates the process of constructing semantic
warehouses. By using the term semantic warehouse we refer to a read-only set of
RDF triples fetched and transformed from different sources that aims at serving a
particular set of query requirements. MatWare automatically fetches contents from
the underlying sources using several access methods (e.g. SPARQL endpoints, HTTP
accessible files, JDBC connections, several file format transformers). The fetched
data are transformed into RDF descriptions using appropriate mappings (Marketakis
et al., 2016), and stored in a RDF triplestore supporting several levels of description
for preserving provenance information. One of its distinctive features, is that it
allows evaluating the connectivity of the semantic warehouse. Connectivity refers to
the degree up to which the contents of the semantic warehouse form a connected
graph that can serve ideally in a correct and complete way the query requirements,
while making evident how each source contributes by using a set of connectivity
metrics. MatWare is a fully configurable tool and can be easily extended using
plugins. For the purposes of GRSF we have extended it with plugins for fetching and
transforming the data from their original formats, plugins for supporting the merging
and dissection steps, as well plugins for publishing the data into the catalogue
supporting both the curation and validation phase, as well as the consumption phase.
334
Fig. 3 The GRSF construction deployment setting
Fig. 3 shows the overall technical deployment for the construction and
maintenance of the GRSF. MatWare is responsible for the activities that construct the
GRSF (as they are described in Section 2.4) and publishing them in the GRSF
Knowledge Base (GRSF KB) and in the GRSF Catalogue. For the latter it exploits
the component Data Catalogue publisher which carries out the necessary activities
for ingesting GRSF records into the CKAN-based Catalogue instance offered by the
D4Science infrastructure. Finally all the above components are controlled and
interacted through the D4Science portal facilities of the GRSF VREs.
4 Evaluation – Discussion
In order to assist the experts during the process of inspecting the GRSF records it
has been decided to keep and publish both the initial records from the database
sources as well as the final GRSF ones. For each one of the final records we preserve
the provenance information about the initial records it has been derived from. This
will allow the experts to quickly identify the problematic sources in cases of
erroneous final records that have to be rejected and undergo a different handling. To
distinguish the initial records from the final ones we used the notion of named graphs
in the GRSF KB; the resources coming from each of the initial data sources have
been added in a particular named graph, and the final records on a different one –
preserving however the links (in terms of URIs) to the corresponding initial resources
that exist in different named graphs. A similar approach has been carried out in the
GRSF data catalogue, where the records are distinguished using groups (e.g. FIRMS
Stock, RAM Stock, GRSF Stock, GRSF Fishery, etc.).
335
Table 1. Summary of the information fetched and integrated into GRSF
FIRMS RAM FishSource GRSF
Stock Records 491 989 873 2,187
Fishery Records 190 - 2,203 7,486
Species 578 264 389 1,204
Water Areas 264 636 302 1,181
Fishing Gears 40 - 59 97
Flag States 69 - 96 163
Assessment Methods 58 74 - 110
Scientific Advices 243 - 326 506
Table 1 summarizes some statistics about the results of the first version of GRSF.
We should note here that only a limited number of merging activities took place.
However, we foresee that in future versions of GRSF we will fully support merging
activities. The table contains the total number of Stock and Fishery records, as well
as the distinct number of particular information like for example the distinct number
of species, water areas and others. What is important to describe here is the high
number of the fishery records in the GRSF database (7,486 records) compared to the
summary of the fishery records in the initial sources (2,393 records in total). This is
due to the merging and dissection processes. More specifically many fishery records
(especially from FishSource) had multiple values for their: (a) target species, (b)
fishing gears, and (c) flag states. These records have been dissected to contain single
values for these fields, a decision that was taken for being compliant with the
guidelines of GRSF to meet the traceability business case, as agreed between the
representatives of the database sources.
The merging process implies the collation of information, thus filling gaps of
knowledge that may occur in the single database sources. In addition integrating the
data sources into a semantic warehouse allows us to create a knowledge graph that
interconnects all the relevant information following the Linked Data principles
(Heath & Bizer, 2011). For instance the three original database sources use different
ways for identifying the targeted species; FIRMS use their common name in English
(e.g. yellowfin tuna), RAM use their scientific Latin name (e.g. Thunnus albacares)
and FishSource use their FAO 3-Alpha code (e.g. YFT). This information is
interconnected in GRSF using the appropriate properties of the top level classes as
shown in Fig. 4. The apparent advantage is that users exploiting GRSF will be able to
find records using any of the above names.
Fig. 4 Different ways for identifying a fish species
336
The semantic warehouse enables querying data from all the underlying sources in
a uniform manner. Table 2 shows the results of the connectivity metrics
(Mountantonakis et al., 2016) (common URIs, literals and triples). GRSF – as a
source – scores the highest value, which justifies that it contains highly connected
and valuable information.
Table 2. Connectivity metrics for GRSF sources as computed by MatWare
Common URIs Common Literals Unique
FIRMS RAM Fish GRSF FIRM RAM Fish GRSF Triples Value
Source S Source (%)
FIRMS 1.0 0.16 % 0.19 % 50.49 % 1.0 3.19 % 7.85 % 45.17 % 82.58 % 0.1087
RAM 1.0 0.14 % 41.71 % 1.0 4.41 % 27.41 % 93.07 % 0.1038
FishSourc 1.0 18.24 % 1.0 8.07 % 97.31 % 0.2011
e
GRSF 1.0 1.0 92.28 % 0.3402
The warehouse can also be exploited as a valuable source of entity names for
improving the quality of automatic semantic annotation of texts and documents, e.g.
by using tools like XLink (Fafalios et al., 2015) that can exploit external SPARQL
endpoints. Moreover one could provide exploratory search services over its contents
by adopting the related approaches that have been developed for RDF datasets (see
Tzitzikas et al., 2016-b for a survey) or by first defining semantic views and then
exploring them through Hippalus, just like it has been done for fish species (Tzitzikas
et al., 2016-c). Finally, it can be exploited for semantically annotating search results
coming from external search systems (Fafalios et al., 2014).
The contents of the GRSF are currently exposed through the D4Science portal.
More specifically two Virtual Research Environments have been created for sharing
knowledge about Stocks and Fisheries, and supports the GRSF business cases. The
first one is for the community of experts for carrying out the curation and validation
activities, and another one is for public use. Fig. 5 shows some indicative screenshots
from the corresponding VREs. By using it, it is possible to search for records by
keyword based search as well as faceting by tags (e.g. commercial species, fishing
area, fishing typology) and groups (e.g. record type, source provider). For each
record the catalogue offers a user-friendly view of its content and cater for
visualizing the time-dependent information associated with it, as well as any other
multimedia resource attached to the record.
337
Fig. 5 Exposing GRSF through the GRSF Virtual Research Environment
As regards the GRSF construction and maintenance we could say that it consists
of three different phases: the design, the implementation and the maintenance phase.
During the design phase the activities for defining the mappings from the original
sources to the targeted format, and the merging and dissection rules are defined.
These processes require human effort and especially they require close collaboration
and interactions with the organizations maintaining the original sources, to make sure
that the semantics of the original data are modelled adequately with respect to the
target models. The output of the design phase is a set of mappings and rules formally
expressed, so that can be used during the implementation phase. The implementation
phase is an automated process that exploits the results of the design phase and
realizes the processes described in Section 2.4. As regards the maintainability of the
GRSF, it is an automated process that can be triggered whenever new data exist in
the original data sources, assuming of course that their structure has been preserved,
otherwise the mappings should also be updated. Finally, we should mention that
during the maintenance phase, any changes made to records by the users (i.e. changes
in the status of a record) are being preserved.
For the particular setting of GRSF, the implementation phase took approximately
47 hours to be completed, and consists of the time for fetching and transforming the
data (~45 minutes), merging and dissecting them (~ 30 minutes) and publishing them
sequentially in the D4Science Data catalogue (~ 45 hours). In future we plan to
parallelize the last subtask for being able to publish multiple records simultaneously.
5 Conclusion – Future Work
In this paper we introduced a process for providing a unified view of several
stocks and fisheries databases, by relying on semantic web technologies and
innovative hybrid data infrastructures. The resulting Global Record of Stocks and
338
Fisheries integrates data from three data sources, and contains more than 9,500
records about stocks and fisheries. It can be seen as a core knowledge base
supporting the collaborative production and maintenance of a comprehensive and
transparent global reference set of stocks and fisheries records. This is accomplished
because of the processes that were applied during the construction, that guarantee the
unique identification of stock and fisheries and the easy access to the information
associated to a particular stock or fishery. In addition, during the validation step, the
experts can validate the information of the GRSF records which also allows them
spotting errors in their original sources, because their provenance is also preserved.
We plan to release newer versions of GRSF periodically that will fully support
the merging activities. Apart from the new versions, we also plan to perform an
inventory for more database sources to be included. Apart from the technical details
of GRSF, there are several issues that are worth further work and research including
(a) support of update operations for specific records in the GRSF data source or the
original sources, (b) parallelization of the construction and maintenance phases for
fastening them, (c) offer advanced discovery services based on spatio-temporal
information, and (d) investigation of whether machine learning techniques could be
exploited for automating or assisting the curation and validation of GRSF records.
Acknowledgments. This work has received funding from the European Union’s
Horizon 2020 research and innovation programme under the BlueBRIDGE project
(Grant agreement No 675680).
References
1. Candela L., Castelli D., Manzi A. and Pagano P. (2014) Realising Virtual
Research Environments by Hybrid Data Infrastructures: the D4Science
Experience. International Symposium on Grids and Clouds (ISGC) Proceedings
of Science PoS (ISGC2014).
2. Assante M., Candela L., Castelli D., Coro G., Lelii L. and Pagano P. (2016)
Virtual research environments as-a-service by gCube. PeerJ Preprints 4:e2511v1
doi:10.7287/peerj.preprints.2511v1.
3. Tzitzikas Y., Minadakis N., Marketakis Y., Fafalios P., Allocca C.,
Mountantonakis M. and Zidianaki I. (2014) MatWare: Constructing and
Exploiting Domain Specific Warehouses by Aggregating Semantic Data. In procs
of the 11th Extended Semantic Web Conference (ESWC'14), Crete, Greece
4. Marketakis Y., Minadakis N., Kondylakis H., Konsolaki K., Samaritakis G.,
Theodoridou M., Flouris G. and Doerr M. (2016) X3ML Mapping Framework for
Information Integration in Cultural Heritage and beyond. Int. Journal on Digital
Libraries, pp 1-19. Springer. DOI 10.1007/s00799-016-0179-1.
5. Tzitzikas Y., Allocca C., Bekiari C., Marketakis Y., Fafalios P., Doerr M.,
Minadakis N., Patkos T. and Candela L. (2016-a) Unifying Heterogeneous and
Distributed Information about Marine Species through the Top Level Ontology
339
MarineTLO. Emerald Group Publishing Limited 50(1)
http://dx.doi.org/10.1108/PROG-10-2014-0072.
6. Hilborn R. and Walters C. (2013) Quantitative fisheries stock assessment:
Choice, Dynamics and Uncertainty. Springer Science & Business Media, ISBN
978-1-4615-3598-0.
7. Heath T. and Bizer C. (2011) Linked Data: Evolving the web into a global data
space. Synthesis lectures on the semantic web: theory and technology. pp. 1-136.
8. Fafalios P., Baritakis M. and Tzitzikas Y. (2015) Exploiting Linked Data for
Open and Configurable Named Entity Extraction. International Journal on
Artificial Intelligence Tools 24(2)
9. Tzitzikas Y., Manolis N. and Papadakos P. (2016-b) Faceted Exploration of
RDF/S Datasets: A Survey, Journal of Intelligent Information Systems (JIIS).
10. Tzitzikas Y., Bailly N., Papadakos P., Minadakis N. and Nikitakis G. (2016-c)
Using preference-enriched faceted search for species identification. IJMSO 11(3)
pp. 165-179.
11. Fafalios P., Papadakos P. and Tzitzikas Y. (2014) Enriching Textual Search
Results at Query Time Using Entity Mining, Linked Data and Link Analysis.
International Journal of Semantic Computing 8(4) pp. 515-544
12. Pham M., Alse S., Knoblock C.A. and Szekely P. (2016) Semantic Labeling: A
domain-independent approach. In proceedings of the 15th International Semantic
Web Conference (ISWC 2016), Japan.
13. Mountantonakis M., Minadakis N., Marketakis Y., Fafalios P. and Tzitzikas Y.
(2016). Quantifying the Connectivity of a Semantic Warehouse and
Understanding its Evolution over Time. International Journal on Semantic Web
and Information Systems (IJSWIS) 12(3), pp. 27-78, DOI:
10.4018/IJSWIS.2016070102
340