AI in Greece: the Case of Research on Linked Geospatial Data
Total Page:16
File Type:pdf, Size:1020Kb
Worldwide AI Worldwide AI AI in Greece: The Case of Research on Linked Geospatial Aust Lorem ipsum alia Data Manolis Koubarakis, George Vouros, Georgios Chalkiadakis, Vassilis Plagianakos, Christos Tjortjis, Ergina Kavallieratou, Greece Dimitris Vrakas, Nikolaos Mavridis, Georgios Petasis, Konstantinos Blekas, Anastasia Krithara ■ We survey the AI research carried out in Greece recently. We concentrate on the case of linked geospatial data, an area with significant practical impor - tance, very interesting research results, rtificial intelligence has been an active research field in and implemented systems developed by a Greek research team. Greece for over 40 years, and there are more than 30 AI Agroups throughout the country covering almost all subareas of AI. A milestone for AI research in Greece came in 1988, when the Hellenic Artificial Intelligence Society (EETN) was founded as a nonprofit scientific organization devoted to organizing and promoting AI research in Greece and abroad. EETN is an affiliated society of the European Association for Artificial Intelligence (EurAI, formerly known as ECCAI). One of the many roles of EETN is the organization of conferences, workshops, summer schools, and other events, such as the Hellenic Conference on Artificial Intelligence (SETN). SETN is Copyright © 2018, Association for the Advancement of Artificial Intelligence. All rights reserved. ISSN 0738-4602 SUMMER 2018 91 Worldwide AI the most prominent forum at which Greek and inter - Berners Lee (2016 Turing Award winner), Jim national AI scientists present original and high-qual - Hendler, and Ora Lassila, found the Institute of Com - ity research on emerging AI topics. The first SETN was puter Science with a team well grounded in KR. That organized in 1996 at the University of Piraeus. Since team went on to become one of the most influential 2002, the conference has been held biennially. The teams in the European semantic web area, especially SETN 2016 conference was co-organized by EETN through the development of RQL, an early query lan - and the Aristotle University of Thessaloniki. Held in guage for the resource description framework (RDF). May of 2016, the conference featured presentations RQL has been greatly influenced by Telos, which had of high-quality AI research work (the acceptance one of the closest representational frameworks to ratio for full papers was below 33 percent). SETN RDF. 3 2016 also attracted international attention, with Manolis Koubarakis was one of the young more than 120 participants from Greece and abroad researchers who worked, under the supervision of joining the event. SETN 2018 1 will take place July 9– John Mylopoulos, on Telos. A few years after earning 15 at the University of Patras. Aside from confer - an MSc in computer science at the University of ences, EETN also organizes a summer school — the Toronto in 1988, Koubarakis returned to Greece and Hellenic Artificial Intelligence Summer School completed his PhD in constraint databases — a (HAISS), which, like SETN, is held on a biennial basis. research area pioneered in 1990 by another Greek, the late Paris Kanellakis, a professor at Brown Uni - Current Lines of versity. Koubarakis then went on to a career in the United Kingdom (Imperial College and University of AI Research in Greece Manchester) and in Greece, first at the Technical Uni - At the time of this writing, AI research is being con - versity of Crete and later at the National and ducted by 31 research groups affiliated with 15 Greek Kapodistrian University of Athens, where he is now universities and research institutes. Many of the based. He was elected a fellow of the European Asso - groups consist of large number of researchers, post - ciation for Artificial Intelligence (EurAI) in 2015. His graduate students, and PhD students working on a Knowledge Representation, Reasoning and Analytics 4 wide variety of topics, 2 including machine learning (KRR&A) group in Athens has done lots of pioneer - and data mining, uncertainty in AI, natural language ing work on linked geospatial data, and the group processing, computer vision, robotics, multiagent constitutes a point of reference for this line of and agent-based systems, constraint satisfaction, research internationally. As such, the remainder of planning and scheduling, and knowledge represen - this article will focus on surveying that work. tation and reasoning. In this article, we will focus on Linked Geospatial Data Work by the the latter, and more specifically on linked geospatial KRR&A Group (2010 – Present) data — an exciting domain with many applications, and one pioneered by the team of Manolis A great deal of geospatial data has recently become Koubarakis. But let us begin first with a quick history available at no charge in Europe and the US, and lesson. there is a strong push for more open geospatial data. This data includes data from volunteer mapping Some History (1968 to Today) efforts such as OpenStreetMap, gazetteers such as Knowledge representation (KR) is an area of research GeoNames, national geographical data made avail - in which Greece has had a strong presence for many able in government data portals, 5 and earth observa - years. This work has been greatly influenced by the tion (EO) data from both the US satellite program work of John Mylopoulos at the University of Toron - Landsat and the Copernicus program of the Euro - to during the 1970s and 1980s. Mylopoulos (who is pean Commission. also an honorary EETN member) supervised a num - Linked data is a research area that studies how one ber of students who ended up in Greek universities can make RDF data available on the web and inter - and spread his influence to younger students enter - connect it with other data with the aim of increasing ing the AI field. Mylopoulos was especially influential its value. In the last few years, linked geospatial data in the work done at the Institute of Computer Sci - has received a lot of attention as researchers and ence, Foundation of Research and Technology – Hel - practitioners have started tapping the wealth of las (ICS-FORTH), based in Heraklion, Crete. Since geospatial information available online. As a result, 1985, Mylopoulos and the researchers at ICS-FORTH the linked open-data cloud has been rapidly popu - have collaborated on a number of ESPRIT projects lated with geospatial data. that have a strong KR component. A core contribu - The KRR&A group was the first international group tion of this research has been the development of the to introduce the linked data paradigm to the EO KR language Telos and its application to engineering domain, and the first such group to develop proto - and cultural informatics. type applications based on transforming earth obser - The arrival of the semantic web in 2001, with the vation data into RDF and combining them with oth - visionary Scientific American paper written by Tim- er kinds of open linked data. Examples of such 92 AI MAGAZINE Worldwide AI Figure 1. The FireHub Service. applications include wildfire monitoring and burnt- them to transform data from many popular geospa - scar mapping, building semantic catalogues for EO tial data formats into RDF. archives, precision farming, change detection using It is often the case that geospatial data is stored in satellite images, real-time maritime situation aware - geospatial relational databases (for example, PostGIS ness, and various environmental services. Probably or SpatialLite) that are not available as linked data. the best example of an application developed by the When these databases are frequently updated or are group is the wildfire monitoring application (figure very large, domain experts are discouraged from 1). The application has been available on the web transforming the data into RDF and storing it in a since 2012, and it was the basis for the operational triple store such as Strabon. To solve this problem, FireHub service deployed at the National Observato - the KRR&A group developed Ontop-spatial, 9 the first ry of Athens. 6 FireHub won the Best Service Chal - geospatial ontology-based data access system. This is lenge in the 2014 Copernicus Masters competition. a very interesting contribution, especially in the light The work of the KRR&A group that concentrated of many published papers on spatial-description log - on the whole lifecycle of linked open EO data and its ics and geospatial databases. Ontop-spatial allows combination with linked geospatial data is shown in domain users to leave their data in geospatial data - figure 2. bases and enables on-the-fly GeoSPARQL-to-SQL The technical challenges of various phases of this translation using ontologies and mappings. The lifecycle have been addressed by the development of experimental evaluation of Ontop-spatial (Bereta and open source tools that represent today’s state of the Koubarakis 2016) has shown that this approach is art in the area of linked geospatial data. Perhaps the not only simpler for the users (because it does not most well-known is Strabon, 7 a spatiotemporal RDF require materialization of data), but also more effi - store. Strabon extends the well-known RDF store cient in terms of query response time when com - Sesame and uses PostgreSQL as the backend spatially pared with state-of-the-art geospatial RDF stores such and temporally enabled DBMS. One of the query lan - as Strabon or even commercial systems. guages supported by Strabon is the Open Geospatial Another of the KRR&A group’s contributions to Consortium standard GeoSPARQL, an extension of the state of the art in visualizing geospatial and tem - the well-known RDF query language SPARQL with poral linked data is development of the tool called vocabulary for querying geospatial data. As shown by Sextant .10 Sextant (figure 3) is a WebGIS for linked the experiments by Garbis, Kyzirakos, and data. It can be used to produce maps by combining Koubarakis (2013), Strabon is currently the most geospatial data from GeoSPARQL endpoints and functional and performant geospatial and temporal well-known GIS file formats.