=Paper=
{{Paper
|id=None
|storemode=property
|title=FCAWarehouse, a Prototype Online Data Repository for FCA
|pdfUrl=https://ceur-ws.org/Vol-1040/paper6.pdf
|volume=Vol-1040
|dblpUrl=https://dblp.org/rec/conf/cubist/OrphanidesG13
}}
==FCAWarehouse, a Prototype Online Data Repository for FCA==
https://ceur-ws.org/Vol-1040/paper6.pdf
FCAWarehouse, a prototype online data
repository for FCA
Constantinos Orphanides and George Georgiou
Conceptual Structures Research Group
Communication and Computing Research Centre
Faculty of Arts, Computing, Engineering and Sciences
Sheffield Hallam University, Sheffield, UK
c.orphanides@shu.ac.uk georgios.georgiou@student.shu.ac.uk
Abstract. This paper presents FCAWarehouse, a prototype online data
repository for FCA. The paper explains the motivation behind the devel-
opment of FCAWarehouse and the features available, such as the ability
to donate datasets and their respective formal contexts, the ability to
generate artificial formal contexts on-the-fly, and how these features are
also available through a set of web-services. The paper concludes by sug-
gesting future work in order to enhance it’s usability.
1 Introduction
In recent years, different means of archiving empirical data have been considered
and implemented to build a space where data collection could be exploited and
analyzed accordingly, if required. The different archiving techniques not only
introduced communities an alternative approach of maintaining the quality and
accessibility of data, but also improved the indexing of available data collections
with proper categorization. An example of such an archiving technique are online
data repositories.
Online data repositories are digital libraries which allow the comprehensive
collection, management and preservation of digital content and the ability to
offer it to targeted user communities [10,7,8]. An example of such a repository is
the UCI Machine Learning Repository, a repository used by the machine learning
community for the empirical analysis of machine learning algorithms [6]. Data
repositories exist for numerous targeted communities; however, a centralized,
public resource of data in FCA formats for the FCA community has not been
implemented to date.
The idea of an FCA online data repository, to provide FCA practitioners with
a resource of public datasets in FCA formats, was proposed in [2] at 2009. The
proposal envisioned the automatic conversion of uploaded traditional datasets
into formal contexts, as well as the ability to generate artificial datasets in CSV
format, to be then converted into a formal context with the user being able
to determine the number of objects, number of attributes and the density. The
automatic conversion of datasets to formal contexts would be handled by a
“Data-to-FCA” converter and the tool FcaStone1 would be used to convert for-
mal contexts from one FCA format to another. Both of these tools would be
incorporated in the repository as components. The architecture of the proposed
repository is shown in Figure 1.
Fig. 1. Proposed FCA Data repository system architecture in [2].
The “Data-to-FCA” converter of the proposed repository was the most com-
plex component to implement with regards to the effort and time needed to de-
velop it. However, since the publication of the aforementioned paper, the “Data-
to-FCA” converter has been developed as a standalone desktop application called
FcaBedrock2 , a formal context creator for FCA with the ability of converting
datasets in various formats to formal contexts in the Burmeister (.cxt) or FIMI
(.dat) formats [5,4]. The formal contexts generated by FcaBedrock can be then
loaded in In-Close3 , a fast formal concept miner, to count the number of for-
mal concepts in a formal context and produce, if necessary, smaller sub-contexts
based on the well-known notion of minimum support [3,1]. Subsequently, the
motivation behind implementing an FCA data repository became stronger and
resulted in the development of a prototype data repository for FCA, named
FCAWarehouse.
1
http://sourceforge.net/projects/fcastone
2
http://sourceforge.net/projects/fcabedrock
3
http://sourceforge.net/projects/inclose
2 FCAWarehouse
FCAWarehouse4 is a prototype data repository for FCA developed as part of a
BSc Computing final year project [9] at Sheffield Hallam University (SHU). It
provides the ability of donating datasets and their respective formal contexts,
browsing and searching datasets, an administration back-end for librarians to
manage donated datasets, creating artificial formal contexts, as well as pro-
viding all of it’s functionality through a set of ReSTful5 web services to make
FCAWarehouse interoperable with third-party FCA tools. A diagram of it’s ar-
chitecture is shown in Figure 2 and an explanation of it’s features is given in the
following sections.
Fig. 2. System architecture of FCAWarehouse.
2.1 Datasets
Each dataset in FCAWarehouse is comprised of metadata such as name, addi-
tional information, original source as well as the original data and it’s respective
formal context (Figure 3).
All of the files and metadata are provided by the donator during the dona-
tion process. In cases where the formal contexts are created using FcaBedrock, a
4
http://www.fcawarehouse.com
5
Representational State Transfer (ReST). http://www.ibm.com/developerworks/
webservices/library/ws-restful/
Fig. 3. Example of a dataset entry in FCAWarehouse (partial screenshot).
Bedrock file (.bed) is also provided which contains the metadata of the conver-
sion so that users can load the original data and the Bedrock file in FcaBedrock
and have a look at, or modify, the parameters and conversion criteria set for
each attribute in the original dataset (Figure 4).
The files donated for each dataset are stored physically on the server, with
the database only holding the URL to each dataset. As the prototype is currently
hosted on a server with limited storage capabilities, all files uploaded for a dataset
during the donation process are automatically compressed as .zip files.
After a user has donated a dataset, the librarians are notified about the new
submission and have the ability of accepting or declining the submitted dataset.
Consequently, only datasets marked as ‘Accepted’ by the librarians are visible
to end-users.
Fig. 4. FCAWarehouse’s Bedrock (.bed) file of the Adult dataset loaded in
FcaBedrock (partial screenshot).
2.2 Artificial formal contexts
FCAWarehouse provides the ability of generating artificial formal contexts in the
Burmeister (.cxt) format by predefining the number of formal objects, number
of formal attributes and the density. Developers of tools such as the formal
concept miners In-Close6 and FCbO7 can use this feature to generate formal
contexts of size of their choice in order to test the limits and efficiency of their
algorithms.
The generated artificial formal contexts are randomly generated by predefin-
ing the number of objects, number of attributes and density of the formal con-
text. The density is defined as a percentage indicating the minimum amount
of crosses (formal attributes) that each row (formal object) should contain in
the formal context. Setting for example number of objects to 10, number of at-
tributes to 10 and density percentage to 70% will result in a formal context with
10 formal objects and 10 formal attributes where at least 70% of each row is
consisted of crosses (Figure 5). Setting the density to 0 will generate a random
artificial formal context with no density criteria.
6
http://sourceforge.net/projects/inclose
7
http://sourceforge.net/projects/fcalgs
Fig. 5. Example of an artificial formal context with 10 objects, 10 formal at-
tributes and density set to 70%.
2.3 Web services
FCAWarehouse implements a set of web services for the interoperability of
FCAWarehouse with third-party FCA applications. At the moment there are
a total of four web services, namely:
– GenerateContext: Accepts as input the number of objects, number of at-
tributes, density percentage and outputs a corresponding artificial Burmeis-
ter formal context in XML.
– GetAllDataSets: Returns all datasets, along with their metadata, in XML.
Only datasets marked as ‘Accepted’ by the librarians are returned.
– Get10MostRecentDatasets: Returns the 10 most recent datasets along
with their metadata, in XML. Only datasets marked as ‘Accepted’ by the
librarians are returned.
– SearchDataset: Accepts a string as input and returns any datasets which
contain the given string in their name, along with their metadata, in XML.
Only datasets marked as ‘Accepted’ by the librarians are returned.
3 Further Work and Conclusion
While useful in it’s current state, FCAWarehouse is still in a prototypical state;
a number of improvements can be implemented to enhance it’s usability. For ex-
ample, the feature of artificially generating formal contexts can be extended to
simulate formal contexts that real datasets would produce. This can be achieved
by defining sets of adjacent columns in the formal context to represent a single
attribute. Assuming, for example, a categorical attribute (with each of it’s values
being a formal attribute in the formal context) and each formal object only being
able to have one of it’s values (quite common in real datasets), will result in a
formal context where no more than one cross will exist in the columns represent-
ing that attribute. The same logic can be applied to various type of attribute:
boolean attributes could be interpreted as single formal attributes in the formal
context and continuous attributes could be grouped using ranges rather than
one formal attribute for each numerical value. In this way, FCAWarehouse can
act as a benchmarker for the comparison of tools and algorithms by providing
citable random data as well as converted real datasets.
A feature which could prove quite useful in the future, with some modi-
fications and adjustments, are FCAWarehouse’s web services. Interesting use-
cases can emerge from this feature; for example, the formal context creator
FcaBedrock could feature a “Create formal context from FCAWarehouse” op-
tion. By using the provided web services, a list of the available datasets could
appear in FcaBedrock. Selecting one of the datasets could automatically down-
load the dataset and start auto detecting its values to create a formal context.
Creating a formal context in FcaBedrock requires inputting a dataset, defining
its metadata and then creating the formal context; considering, however, the
“Generate Context” web service, the ability of creating artificial formal contexts
in FcaBedrock without requiring initial data as input could be made possible
with minimal effort.
Further work also includes the incorporation of the tool FcaStone, to convert
data between FCA and non-FCA formats, in FCAWarehouse to offer the power
of FCA to those currently outside of the FCA community.
The final vision of FCAWarehouse is of an online FCA data repository which
facilitates the creation, conversion and donation of datasets for FCA, providing
a useful collection of real and artificial datasets in a wide variety of FCA and
non-FCA formats to open the way for the wider use of FCA.
References
1. Andrews, S. (2011). In-Close2, a High Performance Formal Concept Miner. In:
Proceedings of the 19th International Conference on Conceptual Structures (ICCS)
2011. LNAI 6828. Berlin: Springer-Verlag. pp. 50–62.
2. Andrews, S. (2009a). Data Conversion and Interoperability for FCA. In: Proceed-
ings of the 4th Conceptual Structures Tool Interoperability Workshop (CS-TIW)
2009, held in conjunction with the 17th International Conference on Conceptual
Structures (ICCS) 2009: “Leveraging Semantic Technologies”. pp. 42–49.
3. Andrews, S. (2009b). In-Close, a Fast Algorithm for Computing Formal Concepts.
In: Rudolph, Dau, Kuznetsov (Eds.): Supplementary Proceedings of ICCS’09,
CEUR WS 483.
4. Andrews, S. and Orphanides, C. (2012). Knowledge Discovery Through Creating
Formal Contexts. In: International Journal of Space-Based and Situated Comput-
ing, 2(2). pp. 123–138.
5. Andrews, S. and Orphanides, C. (2010). FcaBedrock, a Formal Context Creator.
In: “Conceptual Structures: From Information to Intelligence”, Proceedings of the
18th International Conference on Conceptual Structures (ICCS) 2010, LNAI 6208.
pp. 181–184.
6. Bache, K. and Lichman, M. (2013) UCI Machine Learning Repository [http:
//archive.ics.uci.edu/ml]. Irvine, CA: University of California, School of In-
formation and Computer Science.
7. Candela, L., Athanasopoulos, G., Castelli, D., El Raheb, K., Innocenti,
P., Ioannidis, Y., Katifori, A., Nika, A., Vullo, G. and Ross, S. (2011)
The Digital Library Reference Model. Last accessed 01 April 2013 at:
http://bscw.research-infrastructures.eu/pub/bscw.cgi/d222816/D3.2b%
20Digital%20Library%20Reference%20Model.pdf
8. DSpace Cambridge (2012). Digital Repositories. [online]. Last accessed 18 April at
http://www.lib.cam.ac.uk/dataman/pages/repositories.html
9. Georgiou, G. (2013). FCAWarehouse, an Online FCA Data Repository. BSc (Hons)
Computing final year project, Sheffield Hallam University, Sheffield, UK, 2013.
10. Scherle, R., Carrier, S., Greenberg, J., Lapp, H., Thompson, A., Vision, T.
and White, H. (2008). Building Support for a Discipline-Based Data Reposi-
tory. In: Third International Conference on Open Repositories 2008, Southamp-
ton. Last accessed 20 March 2013 at: http://pubs.or08.ecs.soton.ac.uk/35/1/
submission_177.pdf