human-aware sensor network ontology: semantic support for empirical data collection

Post on 08-Feb-2017

502 Views

Category:

Data & Analytics

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Human-Aware Sensor Network Ontology (HASNetO): Semantic Support for

Empirical Data Collection

Paulo Pinheiro1, Deborah McGuinness1, Henrique Santos1,2

1Rensselaer Polytechnic Institute, USA2Universidade de Fortaleza, Brazil

ISWC/LISC, October 2015

Outline

• Capturing Contextual Knowledge• Integration of Empirical Concepts and

Sensor Network Concepts• Provenance Knowledge support for

Contextual Knowledge• HASNetO: The Human-Aware Sensor

Network Ontology • Conclusions

2

Database

Sensornetwork

technician scientist

data user(including scientists)

maintains(deploys,calibrates)

Individual Instrument(s)

measurementdata

measurement Data (e.g., CSV file)

queries

uses

reportsneeds

data flows

interactions

senses

senses

senses

Knowledge Capture

Measurement Time Interval

TimeStamp,AirTemp_C_Avg,RH_Pct_Avg 2015-02-12T09:30:00Z,-4.5,66.582015-02-12T09:45:00Z,-4.372,66.452015-02-12T10:00:00Z,-4.146,65.982015-02-12T10:15:00Z,-4.084,66.222015-02-12T10:30:00Z,-4.251,67.482015-02-12T10:45:00Z,-4.185,69.852015-02-12T11:00:00Z,-4.133,722015-02-12T11:15:00Z,-3.959,70.84…2015-02-12T23:00:00Z,-9.63,77.882015-02-12T23:15:00Z,-10.48,80.82015-02-12T23:30:00Z,-10.96,822015-02-12T23:45:00Z,-10.1,80.7

t

A Comma-Separated Value (CSV) dataset:

February 12, 2015, 9:30AM

February 12, 2015, 11:45PM

Temporal Contextual Diff

t

Configuration

Deployment

SensorCalibration

InfrastructureAcquisition

t

February 12, 2015, 9:30AM

February 12, 2015, 11:45PM

Data usage

Full Extent of Contextual Knowledge Scope

6

timespaceagentstrust

“typical” measurement scope

Selected Observation and Sensor Network Ontologies

• Sensor Network Knowledge– Needed to describe the infrastructure of a

sensor network, and the use of sensor network components in the generation of datasets

• Observation Knowledge– Needed to describe observations and their

measurements. Measurements need to be characterized in terms of physical entities, entity characteristics, units, and values

Observation ConceptsIn our measurements, observation concepts are either OBOE concepts or OBOE-derived concepts.

The thing that one is observing is an entity, e.g.,’air’.

Things that are observed, however, cannot be measured. For example, how can one measure ‘air’? A characteristic is a measurable property of an entity, e.g., air temperature.

An observation is a collection of measurements of entity’s characteristics.

Each measurement has a value, e.g, ’45’, and a standard unit, e.g., ‘Celsius’.

oboe:Entity

oboe:Observation

of-entity11

hasneto:DataCollection

oboe:Measurement

oboe:Standard

oboe:Characteristic

oboe:Value

of-characteristic

hasneto:hasMeasurement

uses-standard

has-characteristic

has-characteristic-value

has-standard-value

has-value

hasneto:hasContext

11

*

1

1

1

1

1

1

*

*

*

*

*

*

Sensor Network ConceptsIn the Jefferson Project, sensor network concepts are either Virtual Solar-Terrestrial Observatory (VSTO) concepts or VSTO-derived concepts.

Instruments and their detectors are used to perform measurements.

Instruments, however, can only perform measurements during a deployment at a given platform, e.g., tower, plane, person, buoy

vstoi:Detector

vstoi:Instrument

vstoi:Platform

hasneto:Sensing

Perspective

oboe:Characteristic

oboe:Entity

vstoi:Detachable

Detector

vstoi:AttachedDetector

* *

*

1

0..1*

hasPerspectiveCharacteristic

perspectiveOf

Selected Provenance Ontology

Provenance Knowledge is needed to contextualize VTSO deployments and OBOE observations

– “Who deployed an instrument?” – “When was the instrument deployed?” – “How many times instrument parameters

changed during deployment?” – “What was the value of each parameter

during a given observation?”

W3C PROV Concepts

Provenance concepts are W3C PROV concepts.

Provenance-Level Integration

• Provenance provides contextual high-level integration of observation and sensor network concepts

• Integration also occurs in terms of information flow allowing full accountability of measurements in the context of sensor network components and configurations

12

prov:Activity

hasneto:DataCollection

vstoi:Deployment

xsd:dateTime

xsd:dateTime

hasDataCollection

1*

prov:Agent

prov:Entity

usedwasGeneratedBy

wasAttributeTo

wasAssociatedWith

actedOnBehalfOf

wasDerivedFrom

startedAtTime

endedAtTime

The Human-Aware Sensor Network Ontology

vstoi:Detector

vstoi:Instrument

vstoi:Platform

hasneto:Sensing

Perspective

oboe:Characteristic

oboe:Entity

vstoi:Detachable

Detector

vstoi:AttachedDetector

*

*

*1

0..1

*hasPerspectiveCharacteristic

perspectiveOf

prov:Activity

hasneto:DataCollection

vstoi:Deployment

xsd:dateTime

xsd:dateTime

hasDataCollection

1*

prov:Agent

wasAssociatedWithstartedAtTime

endedAtTime

1

1

*

**

*

oboe:Measurement

of-characteristic

hasneto:hasMeasurement 1

1

*

*

Metadata in Action

14

Mouse over

Combining Data and Metadata

15

Mouse over

Mouse over

Metadata

based

facete

d searc

h

Measurement metadata

Metadata about the metadata

Conclusions

• HASNetO was briefly presented along with its support for describing sensor networks

• OBOE and VSTO provide concepts required for encoding observation and sensor network metadata

• Neither OBOE and VSTO provide concepts for describing contextual knowledge about deployments and observations

16

HASNetO provides a comprehensive integrated set of concepts for capturing sensor network measurements along with contextual knowledge about these measurements

• Extra

17

SPARQL Queries Against HASNetO

• Question in English:“List detectors currently deployed with instrument vaisalaAW310-SN000000 and the physical characteristics measured by these detectors”

• W3C SPARQL query (a translation of the question above):select ?detector ?characteristic ?platform where {?deployment a Deployment>. ?deployment vsto:hasInstrument kb:vaisalaAW310-SN000000. ?platform vsto:hasDeployment ?deployment. ?deployment hasneto:hasDetector ?detector. ?detector oboe:detectsCharacteristic ?characteristic. }

• Query Result:+----------------+-------------------+--------------------+

| detector | characteristic | platform |+----------------+-------------------+--------------------+ | Vaisala WMT52 | windSpeed | towerDomeIsland |+----------------+-------------------+--------------------+

18

Example of a HASNetO Knowledge Base*

19

:obs1 a oboe:Observation; oboe:ofEntity oboe:air; prov:startedAtTime "2014-02-11T01:01:01Z"^^xsd:dateTime;

prov:endedAtTime "2014-02-12T01:01:01Z"^^xsd:dateTime; .  :dp1 a vsto:Deployment;

vsto:hasInstrument :vaisalaAW310-SN000000; hasneto:hasDetector :vaisalaWMT52-SN000000;

hasneto:hasObservation :obs1;prov:startedAtTime "2014-02-10T01:01:01Z"^^xsd:dateTime; prov:endedAtTime "2014-02-17T01:20:02Z"^^xsd:dateTime; .

 :genericTower vsto:hasDeployment :dp1; .  :dset1 a vsto:Dataset;

prov:wasAttributedTo :vaisalaAW310; prov:wasGeneratedBy :obs1; .

*The knowledge base fragment above is represented in W3C Turtle.

Knowledge About Sensor Network Operation

• Knowledge about sensor networks, however, can rarely be inferred from sensor data themselves.

• The lack of contextual knowledge about sensor data can render them useless.

Knowledge about sensor networks is as important as data captured by sensor networks, and sensor network metadata is as important as sensor data

21

Human-Aware Data Acquisition Framework

• Two locations: • Darrin Fresh Water

Institute (DFWI) at Lake George, NY and

• data processing site in Troy, NY

• Wireless network used to communicate with sensors

• Relational database for data management and RDF triple store for metadata management

Future Steps

• We will keep refining the HASNetO vocabulary and testing it over a constantly growing HASNetO-based knowledge base

• We are in the process of integrating HASNetO into the HAScO (Human-Aware Science Ontology) to accommodate contextual knowledge beyond observation data to include simulation data and experimental data

22

top related