Dimitris Kotzinos

Linked open GeoData management in the cloud

The need to better integrate and link various isolated data sources on the web has been widely recognized and is tackled by the Linked Open Data (LOD) initiative. One of the problems to address is the issue of publishing and subsequently exploiting the data as LOD, due to reasons of data size and performance of the respective queries and to the publication complexity. This work addresses the size and performance issues by adapting the cloud as a hosting platform for LOD publication services so as to exploit its scalability and elasticity capabilities. The publication complexity issue is addressed by proposing a Linked Open Data-as-a-Service approach offering an integrated service based API for (semi)automatic publication of relational data as LOD and subsequent querying and updating capabilities.

Analysis of Quantitative Profiles of GI Education: towards an Analytical Basis for EduMapping

There is an ongoing discussion among the members of the GI educational community about the possibility to find a common way to describe a course taught as part of a GI curriculum (anywhere in the world) with the final goal of being able to automatically identify similar courses and define their equivalence. EduMapping is such an initiative that started recently, which used the BoK concepts as its basic labeling scheme. Based on this work we extended the analysis provided by the EduMapping initiative by suggesting and applying an analytical method that is capable of clustering the courses into classes based on (dis)similarity metrics, which are in turn calculated based on the course assessments done by their instructors using the BoK concepts. In this paper, we present and discuss the preliminary results obtained while applying the suggested method on the EduMapping data. We also provide some pointers for further research in an area that has very few contributions so far.

A Cloud-Based, Geospatial Linked Data Management System

The Web has been evolving to a sink of disparate informa- tion sources which are totally isolated from each other. The technology of Linked Data (LD) promises to connect such information sources in order to enable their better exploitation by humans or automated pro- grams. While various LD management systems have been proposed, only few of them are able to handle geospatial data which are becoming quite popular nowadays and lead to the creation of large geospatial footprints. However, none of the few systems that support Linked Open Geospa- tial Data is able to scale well to handle the increasing load from user queries. In addition, the publishing of geospatial LD also becomes quite advantageous due to complexity reasons. To this end, this article pro- poses a novel, cloud-based geospatial LD management system which can scale out or scale in according to the incoming load in order to serve the respective user requests with the appropriate service level. On top of this system lies a LD-as-a-service offering which abstracts away the user from any LD publishing complexities and provides all the appro- priate functionality for enabling a full LD management. We also study and propose architectural solutions for the distributed update problem. The proposed system is evaluated under heavy load scenarios and the results show that the respective improvement in performance incurred is quite satisfactory and that the scaling actions are performed at the appropriate time points.

Cloud Based Processing Services Based on Linked Data

Cloud computing is providing a computing infrastructure to facilitate storage and processing of massive amounts of information (Big Data). Processing of massive datasets becomes more and more important since the data becoming available to us increase every day in volume, variety, speed of change and (potentially) quality. Processing these data becomes more and more difficult under traditional computing platforms since we need the ability to compute and scale at the same time.Under this context, for this work we describe the design and implementation of a responsive and user driven processing service. This is a geoprocessing service that operates on geospatial datasets and provides geostatistical interpolation (a specific variant called Kriging). This service is based on existing service implementation standards in the geospatial domain (namely WPS standard from OGC). Additionally our service can query and retrieve information that is integrated following the Linked Open Data (LOD) initiative. This is a unique capability that allows the service to rely on data, besides the existing ones and the ones provided by the user, that can be retrieved from the integrated information space that is being built on the web. In this paper we present the design and implementation of the service on a Linked Data store and discuss capabilities, issues and future research.

A Methodological Framework for Statistical Analysis of Social Text Streams

Social media are one of the main contributors of user generated content; providing vast amounts of data in daily basis, covering a wide range of topics, interests and events. In order to identify and link meaningful and relevant information, clustering algorithms have been used to partition the user generated content. We have identified though that these algorithms exhibit various shortcomings when they have to deal with social media textual information, which is dynamic and streaming in nature. Thus we explore the idea to estimate the algorithms’ parameters based on observations on the clusters’ properties’ (like the centroid, shape and density) evolution. By experimenting with the clusters’ properties, we propose a methodological framework that detects the evolution of the clusters’ centroid, shape and density and explores their role in parameters’ estimation.