Markus Stocker

SPARQL basic graph pattern optimization using selectivity estimation

In this paper, we formalize the problem of Basic Graph Pattern (BGP) optimization for SPARQL queries and main memory graph implementations of RDF data. We define and analyze the characteristics of heuristics for selectivity-based static BGP optimization. The heuristics range from simple triple pattern variable counting to more sophisticated selectivity estimation techniques. Customized summary statistics for RDF data enable the selectivity estimation of joined triple patterns and the development of efficient heuristics. Using the Lehigh University Benchmark (LUBM), we evaluate the performance of the heuristics for the queries provided by the LUBM and discuss some of them in more details.

The fundamentals of iSPARQL: a virtual triple approach for similarity-based semantic web tasks

This research explores three SPARQL-based techniques to solve Semantic Web tasks that often require similarity measures, such as semantic data integration, ontology mapping, and Semantic Web service matchmaking. Our aim is to see how far it is possible to integrate customized similarity functions (CSF) into SPARQL to achieve good results for these tasks. Our first approach exploits virtual triples calling property functions to establish virtual relations among resources under comparison; the second approach uses extension functions to filter out resources that do not meet the requested similarity criteria; finally, our third technique applies new solution modifiers to post-process a SPARQL solution sequence. The semantics of the three approaches are formally elaborated and discussed. We close the paper with a demonstration of the usefulness of our iSPARQL framework in the context of a data integration and an ontology mapping experiment.

Semantic Process Retrieval with iSPARQL

The vision of semantic business processes is to enable the integration and inter-operability of business processes across organizational boundaries. Since different organizations model their processes differently, the discovery and retrieval of similar semantic business processes is necessary in order to foster inter-organizational collaborations. This paper presents our approach of using iSPARQL --- our imprecise query engine based on iSPARQL --- to query the OWL MIT Process Handbook --- a large collection of over 5000 semantic business processes. We particularly show how easy it is to use iSPARQL to perform the presented process retrieval task. Furthermore, since choosing the best performing similarity strategy is a non-trivial, data-, and context-dependent task, we evaluate the performance of three simple and two human-engineered similarity strategies. In addition, we conduct machine learning experiments to learn similarity measures showing that complementary information contained in the different notions of similarity strategies provide a very high retrieval accuracy. Our preliminary results indicate that iSPARQL is indeed useful for extending the reach of queries and that it, therefore, is an enabler for inter- and intra-organizational collaborations.

Situational Knowledge Representation for Traffic Observed by a Pavement Vibration Sensor Network

Information systems that build on sensor networks often process data produced by measuring physical properties. These data can serve in the acquisition of knowledge for real-world situations that are of interest to information services and, ultimately, to people. Such systems face a common challenge, namely the considerable gap between the data produced by measurement and the abstract terminology used to describe real-world situations. We present and discuss the architecture of a software system that utilizes sensor data, digital signal processing, machine learning, and knowledge representation and reasoning to acquire, represent, and infer knowledge about real-world situations observable by a sensor network. We demonstrate the application of the system to vehicle detection and classification by measurement of road pavement vibration. Thus, real-world situations involve vehicles and information for their type, speed, and driving direction.

Representing situational knowledge acquired from sensor data for atmospheric phenomena

Abstract A recurrent problem in applications that build on environmental sensor networks is that of sensor data organization and interpretation. Organization focuses on, for instance, resolving the syntactic and semantic heterogeneity of sensor data. The distinguishing factor between organization and interpretation is the abstraction from sensor data with information acquired from sensor data. Such information may be situational knowledge for environmental phenomena. We discuss a generic software framework for the organization and interpretation of sensor data and demonstrate its application to data of a large scale sensor network for the monitoring of atmospheric phenomena. The results show that software support for the organization and interpretation of sensor data is valuable to scientists in scientific computing workflows. Explicitly represented situational knowledge is also useful to client software systems as it can be queried, integrated, reasoned, visualized, or annotated.

Making Sense of Sensor Data Using Ontology: A Discussion for Residential Building Monitoring

We illustrate the application of automated representation of knowledge acquired from sensor network data to quality of life services. Specifically, for a sensor network used to monitor a residential building we acquire knowledge about events of interest to occupants and represent such knowledge in ontology. An event of particular interest to quality of life which we discuss is ‘unhealthy’ exposure to carbon monoxide. Hence, we aim at reducing the considerable gap between raw sensor data and abstract domain terminology. Our results support the claim that computational techniques in signal processing, machine learning, and ontology engineering are important elements to systems that make use of environmental sensing, including systems for quality of life information services.

The Relevance of Measurement Data in Environmental Ontology Learning

Ontology has become increasingly important to software systems. The aim of ontology learning is to ease one of the major problems in ontology engineering, i.e. the cost of ontology construction. Much of the effort within the ontology learning community has focused on learning from text collections. However, environmental domains often deal with numerical measurement data and, therefore, rely on methods and tools for learning beyond text. We discuss this characteristic using two relations of an ontology for lakes. Specifically, we learn a threshold value from numerical measurement data for ontological rules that classify lakes according to nutrient status. We describe our methodology, highlight the cyclical interaction between data mining and ontologies, and note that the numerical value for lake nutrient status is specific to a spatial and temporal context. The use case suggests that learning from numerical measurement data is a research area relevant to environmental software systems.

Detection and Classification of Vehicles by Measurement of Road-Pavement Vibration and by Means of Supervised Machine Learning

Road vehicle detection and, to a lesser extent, classification have received considerable attention, in particular for the purpose of traffic monitoring by transportation authorities. A multitude of sensors and systems have been developed to assist people in traffic monitoring. Camera-based systems have enjoyed wide adoption over the last decade, partially substituting for more traditional techniques. Methods based on road-pavement vibration are not as common as camera-based systems. However, vibration sensors may be of interest when sensors must be out of sight and insensitive to environmental conditions, such as fog. We present and discuss our work on detection and classification of vehicles by measurement of road-pavement vibration and by means of supervised machine learning. We describe the entire processing chain from sensor data acquisition to vehicle classification and discuss our results for the task of vehicle detection and the task of vehicle classification separately. Using data for a single vibration sensor, our results show a performance ranging between 94% and near 100% for the detection task (1340 samples) and between 43% and 86% for the classification task (experiment specific, between 454 and 1243 samples).

An alternative approach to transverse and profile terrain curvature

Terrain curvature is one of the most important parameters of land surface topography. Well-established methods used in its measurement compute an index of plan or profile curvature for every single cell of a digital elevation model (DEM). The interpretation of these outputs may be delicate, especially when selected locations have to be analyzed. Furthermore, they involve a high level of simplification, contrasting with the complex and multiscalar nature of the surface curvature itself. In this paper, we present a new method to assess vertical transverse and profile curvature combining real-scale visualization and the possibility to measure these two terrain derivatives over a large range of scales. To this purpose, we implemented a GIS tool that extracts longitudinal and transverse elevation profiles from a high-resolution DEM. The performance of our approach was compared with some of the most commonly used methods (ArcMap, Redlands, CA, USA; ArcSIE, Landserf) by analyzing the terrain curvature around charcoal production sites in southern Switzerland. The different methods produced comparable results. While conventional methods quickly summarize terrain curvature in the form of a matrix of values, they involve a loss of information. The advantage of the new method lies in the possibility to measure and visualize the shape and size of the curvature, and to obtain a realistic representation of the average curvature for different subsets of spatial points. Moreover, the new method makes it possible to control the conditions in which the index of curvature is calculated.

Knowledge-based environmental research infrastructure: moving beyond data

Over the past decades, sensor networks have been deployed around the world to monitor over time and space a large number of properties appertaining to various environmental phenomena. A popular example is the monitoring of particulate matter and gases in ambient air undertaken, for instance, to assess air quality and inform decision makers and the public. Such infrastructure can generate large amounts of data, which must be processed to derive useful information. Infrastructure may be for environmental research, specifically. In order to reduce duplication and improve interoperability, efforts have been initiated more recently that aim at abstract architectural descriptions of infrastructure that supports the acquisition, curation, access, and processing of measurement and observation data. The ENVRI Reference Model (ENVRI-RM) is an example for an abstract architectural description of infrastructure tailored for environmental research. We briefly summarize ENVRI-RM and provide an overview of its subsystems, functionality, and viewpoints. We highlight that its primary focus is on the data life-cycle in environmental research infrastructure. As our contribution, weextend ENVRI-RM with functionality for the acquisition of knowledge from data, and the curation, access, and processing of knowledge. The extension, which we name +K, aims at addressing the knowledge life-cycle in environmental research infrastructure. We present the +K subsystems and functionality, and discuss the extension from ENVRI-RM viewpoints. We argue that the +K extension can be superimposed on ENVRI-RM to form the ENVRI-RM+K model for the ‘archetypical’ knowledge-based environmental research infrastructure that addresses both data and knowledge life-cycles. We demonstrate the application of the extension to a concrete use case in aerosol science.