Today we have had a whole day workshop at the ISWC2012 conference on the topic of Semantic Sensor Networks. It is already the 5th workshop that has been held in this area in the last years, and was at the core of the decision of starting a W3C incubator group on Semantic Sensor Networks, in which I participated with another OEGer, Raúl García, which produced interesting results like the SSN Ontology and a semantic markup proposal for sensor data.
Before you start asking where the papers are, you can get them from CEUR proceedings Vol 904, and several of the presentations will be made available in around two weeks at Videolectures. And also before I forget, the work done by Kerry Taylor and Cory Henson on the organisation of the workshop has been great (great people to work with).
And I will start with the invited talk by David de Roure, even if this came in the end of the workshop, talking about the relationship between semantic sensor networks, science, social aspects and..., music... Some interesting projects in this area could be the Semantic Media Network or Social Machines.David presented some examples of work that he has been involved in. For instance, the Linked Data repositories generated in the SALAMI project, related to the area of music information retrieval. And obviously referred to workflows (e.g., a paper titled Reuse, Remix, Repeat: the Workflows of MIR), so as to deal with the relationship between data and methods to deal with such data, and to ensure reproducibility, reuse and sharing. And from there to Research Objects, such as those defined in the Wf4Ever project.
On the social area, and how this can help music, Dave presented how a crowdsourcing approach can be useful to obtain data from old music scores.
As a summary Dave presented two case studies: one on ICT-enabled manufacturing, where sensors can be used throughout the whole production process, and where data can also come from discourse in collaborative engineering processes; and the another one on end-to-end semantics in the music production and consumption process, from instruments (acting as sensors) to the music consumers, with the scores and MP3 files acting as important parts of the aforementioned Research Objects.
A final set of comments:
- Do we have examples of end-to-end semantics in SSN?
- How do we share the methods for processing sensor data? Provenance, trust, reproducibility
- Are we considering SSN as sociotechnical systems? what are our social objects in SSN, what is the social life of objects, can we use sensors as citizens, what are the social machines in SSN?
- What can we do in concert with science and music?
- Can we build something similar to Music Information Retrieval to advance the work?
Once I have gone through this inspiring talk, I will try to summarise the main topics discussed during the workshop. My view is that the following topics were the most salient ones:
- Applications and patterns that use, reuse and extend the SSN Ontology. Much of this work is not only useful for our Semantic Sensor Network community, but also for the ontology engineering community.
- Technology needed (query and event processing systems, sensor middleware and platforms) to provide support for semantic sensor networks.
- Characterisation of sensor data streams.
I will now provide some brief descriptions of the main highlights of most of the presentations (please note that they are not in order of presentation, but according to my previous characterisation).
1. Applications and patterns that use, reuse and extend the SSN Ontology
Laurent Lefort, from CSIRO, presented the award-wining (btw, thanks to the SPITFIRE project for sponsoring it) paper A Linked Sensor Data Cube for a 100 Year Homogenised Daily Temperature Dataset, describing the data cubes that have been made available at http://lab.environment.data.gov.au/, reusing the SSN Ontology and the RDF Datacube vocabulary. Many of the problems that he was explaining remind me of the ones that we had when building our http://aemet.linkeddata.es/ site, which used data similar to that used here from the Australian BoM (Bureau of Meteorology) weather stations and which was made available in the now not-so-open AEMET FTP site. Some of these problems are the changes in the sites where weather stations are, the changes in numbers, the sensors, the procedure changes (hours of observations), etc. I recommend checking the mashup. The data is also accesible through the Linked Data API (for example, http://lab.environment.data.gov.au/data/acorn/climate/slice/station). Great work, Laurent, and good discussion during coffee afterwards.
Edna Ruckhaus, from Universidad Simón Bolívar, was presenting some joint work with our Ontology Engineering Group on a case study application with bike sharing systems. The resulting mashup is available at http://transporte.linkeddata.es/. This paper is not only about the description of how the application was built, but also about lessons learned in the process (e.g., difficulties found when extending the SSN Ontology on the treatment of properties). Besides, it showed how R2RML mappings can be used to generate static metadata about sensors, as well as generating data streams, with the same set of mappings.
Rimel Bendadouche, from Irstea, presented the paper Extension of the Semantic Sensor Network Ontology for Wireless Sensor Networks: The Stimulus-WSNnode-Communication Pattern, focused on the topic of flood modelling. Everybody in the audience agreed that this was a nice pattern to consider and that it would be good to have it discussed in the SSN community group.
Mikel Emaldi presented the work behind BizkaiSense (check the paper here), in the context of the Open Data Euskadi initiative, especially focusing on the ontologies that had to be taken and extended from the SWEET suite of ontologies or for units of measurement. Now they are starting to work on solid waste.
Auriol Degbelo, from Münster, presented a data quality extension to the SSN Ontology.
Another interesting sensor registration demo was given by Pramod Anantharam, from Wright State University,
2. Enabling technology
Mikko Rinne, from the Aalto University, presented his paper titled SPARQL-Based Applications for RDF-Encoded Sensor Data. Even if the title may suggest that the presentation would be mainly about applications, Mikko's presentation focused mainly on the implementation of the INSTANS event processing platform, based on SPARQL and the use of the RETE algorithm. Two implementations exist (on SCALA and LISP, yes, that old language that is still being used in many applications, and that according to Mikko is much faster). Following Mikko's presentation, I am really willing to test it (looking forward to having on their website the LISP version available).
Luka Bradesko, from JSI, presented a rule-based framework and tool for acquiring semantic metadata for sensors.
Ruben Verborgh, a PhD student from Ghent University, presented the paper Functional Composition of Sensor Web APIs, describing how sensors, described as REST Web APIs, can be automatically composed with generic reasoners in a short amount of time (a nice example used in the presentation was on looking for a restaurant to have dinner, and sitting outside in case that it will not rain). More information is available at http://restdesc.org/.
Kia Teymourian, from Freie Universität Berlin, presented the paper Semantic Processing of Sensor Event Stream by Using External Knowledge Bases. This work addresses an important aspect of how to enrich data streams with additional external knowledge bases in order to detect events in those data streams, using a description logic reasoner. It showed how SPARQL queries can be intertwined with event processing queries for a more complex event processing that takes into account such external knowledge bases.
3. Sensor data characterisation
Jean Paul Calbimonte, from my research group, presented some joint effort between our group and EPFL on the characterisation of sensor data streams. Starting from the assumption that especially in citizen sensing projects the metadata that is given by data stream owners is usually weak, it may possible to look at the data streams and generate the SSN Ontology-related metadata that is needed in order to understand better the features of interest and properties that these data streams are measuring.In summary, a very lively workshop with lots of new ideas and the confirmation that the work that the Semantic Sensor Community has done in the past is currently being used. I hope that this summary is useful for my readers, and I hope that next year we will be holding our 6th Workshop. See you in Sydney, home of the ISWC2013 conference.
