Signi cant e orts have been made to understand and document knowledge related to scienti c measurements. Many of those efforts resulted in one or more high-quality ontologies that describe some aspects of scienti c measurements, but not in a comprehensive and coherently integrated manner. For instance, we note that many of these high-quality ontologies are not properly aligned, and more challenging, that they have di erent and often con icting concepts and approaches for encoding knowledge about empirical measurements. As a result of this lack of an integrated view, it is often challenging for scientists to determine whether any two scienti c measurements were taken in semantically compatible manners, thus making it di cult to decide whether measurements should be analyzed in combination or not. In this paper, we present the Human-Aware Sensor Network Ontology that is a comprehensive alignment and integration of a sensing infrastructure ontology and a provenance ontology. HASNetO has been under development for more than one year, and has been reviewed, shared and used by multiple scienti c communities. The ontology has been in use to support the data management of a number of large-scale ecological monitoring activities (observations) and empirical experiments.
Scienti c communities are experiencing a signi cant increase in the availability of empirical data due to the falling cost of sensors along with the growing ease of sensor deployment and with sensor data distribution over the internet. The same communities are also experiencing increasing pressure from a variety of stakeholders to see their empirical data consolidated, analyzed, and used to explain a broad range of unanswered scienti c problems. However, this consolidation and analysis presents challenges since scientists are not yet fully equipped to understand the quality and semantics of scienti c measurements with the data and often limited annotations typically available today. Many voice a strong need for a comprehensive vocabulary capable of encoding and supporting systematic understanding of metadata about empirical data, which would enable sound integration of empirical data.
We present the Human-Aware Sensor Network Ontology (HASNetO) that is a comprehensive alignment and integration of well-established ontologies for encoding scienti c sensing infrastructures, scienti c observations, and provenance. The integrated ontology is available at http://hadatac.org/ont/hasneto. Supporting ontologies for HASNetO and previous versions of HASNetO can also be found at http://hadatac.org. A comprehensive infrastructure for managing HASNetO-based knowledge bases, which is not discussed in this paper, is available at https://github.com/paulopinheiro1234/hadatac. One of the immediate bene ts of HASNetO is its capability of describing comprehensive knowledge graphs about empirical data.
We have used this graph to systematically annotate and amplify the relevance
of scienti c measurements stored in database systems, in support of three major
projects: Je erson Project [
The rest of this paper is organized as follows. In Section 2, we use a diagram to discuss a typical scenario where empirical data is generated and managed. Section 3 presents a categorization of knowledge related to scienti c measurements that is often described as measurement metadata. In Section 4, we introduce the Human-Aware Sensor Network Ontology that provides concepts and relationships used to encode the knowledge discussed in Section 3. In Section 5, we compare our work on HASNetO with other initiatives. A more comprehensive discussion about the current impact of our HASNetO work including future work is described in Section 6. Finally, we summarize our work in Section 7. 2
Empirical data are often collected with the use of instruments that are manually operated by scientists, and sensor networks that are automatically operated but that are still deployed, calibrated and maintained manually.
The three faces depicted in Fig. 1 represent human roles (or just roles) in a data collection scenario. The Scientist role is connected to a Technician role representing the fact that scientists interact with technicians to communicate their needs in terms of how sensor networks are required to be set up and maintained. Human roles in Figure 1 can be performed by any combination of people and roles. 3 http://www.case.rpi.edu/page/academics.php
In Figure 1, Diamond Category \1" represents knowledge about available measurement infrastructure. Scientists conceptually understand the con guration and capabilities of measurement infrastructures, including which instruments and detectors are available to them, which platforms these instruments can be or are deployed to, where stationary platforms are located, which paths are taken by mobile platforms, and what physical, chemical, biological and sociological properties the sensors are capable of measuring. Assuming that the knowledge about instruments and sensor networks may a ect empirical data understanding, scientists are expected to share their measurement infrastructure knowledge by encoding such knowledge as measurement data's metadata.
In Figure 1, Diamond Category \2" represents knowledge about a broad range of human interventions that may a ect the quality of measurement data. Knowledge about measurement infrastructure is not nearly enough to explain data generated by instruments in such infrastructures. For instance, many are the factors/events that may a ect the way measurements are performed, which are not included in the knowledge about the measurement infrastructure itself. When scientists are operating scienti c instruments in isolation, it is evident the importance of documenting how the instruments were operated. More challenging is the process of explaining human interventions in sensor networks, which is often regarded as an automated infrastructure for the collection of scienti c data. For example, a badly deployed instrument, e.g., an instrument that is not properly attached to the surface of the deploying platform, can create measurements that are o by a xed amount, or even worse, that may not be able to execute any measurement, e.g., because the chord providing power to the instrument is not properly connected. 3.3
In Figure 1, Diamond Category \3" represents knowledge about what has been measured, and how the measurement has been represented in terms of units. When measurements occur, these are measurements of physical, chemical, biological, cultural, and social properties of so-called entities of interests. For example, using Air as an entity of interest, we can say that the air temperature is a physical property of air and that the CO2 concentration of the air is a chemical property of the same entity of interest. For data understanding, it is important for one to know the properties are that are being measured, e.g., temperature and CO2 concentration, and what entities of interest are behind these properties, e.g., air. Moreover, it is important to understand the unit used to represent the measurements and the semantic context, (e.g., air is `outside air' as opposed of air inside of a room, a lab or a shelter) that may a ect, for instance, the actual measurement of both air temperature and air concentration of CO2. 3.4
In Figure 1, Diamond Category \4" represents knowledge about the provenance of both human interventions, as well as of each measurement. For each measurement, it is important to know when and where the measurement was done. What was the combination of sensing devices used to support the measurement? Was any con guration parameter provided to the sensing devices to allow the devices to operate the way they were operating at the time the measurements were done?
HASNetO aims to provide the concepts and vocabularies needed to encode
empirical data's metadata as identi ed and described in Section 3. HASNetO is
built on top of three ontologies that were integrated and extended under the
single name of HASNetO: The Extensible Observation Ontology (OBOE) [
HASNetO contains content related to sensor networks although it does not have a Sensor concept. We observe that the term sensor is used to refer to detectors, instruments, and often to combinations of detectors and instruments. To avoid further confusion, HASNetO advocates for the use of the terms detectors and instruments knowing that it may be di cult to perceive which part of a device is a detector (or detectors) and which part is an instrument. For instance, a thermometer may include an embedded, non-detachable detector. HASNetO breaks down the elements of measuring infrastructures into three categories, as shown in Fig. 2.
{ vstoi:Platform: An object that keeps the instrument in a speci c location to ensure that it is recording data about the selected location. A platform may also provide overhead services, such as providing power to the instrument and a data connection. Sometimes a platform is mobile like a plane or a person, or stationary like a tower of a weather station. { vstoi:Instrument : n object that receives sensed signals from detectors and processes these signals into numerical values. For example, consider a tipping bucket rain gauge. Inside the tipping bucket rain gauge is a magnet-based detector that detects when the bucket tips. However, in order for this signal to be meaningful, the detector needs a bucket with a known diameter, a funnel to direct water into the bucket, etc. Together, these make up the instrument. { vstoi:Detector : An object that it is capable of sensing environmental properties by collecting physical signals about these properties, translating these physical signals into (most often electrical) signals, and forwarding these electrical signals to instruments. Transducer is another name for detector. Detector metadata are collected because detectors may be interchangeable, that is they can be removed from one instrument and plugged into another. { vstoi:Deployment : An activity of physically deploying an instrument and its attached detectors to a platform. This activity indicates that a single instrument is ready to start collecting data. 4 The vstoi namespace refers to the instrument portion of the VSTO ontology family.
For more sophisticated devices, detectors and instruments are sometimes available as distinct hardware components, and thus easier to be mapped into HASNetO concepts. For ordinary instruments, it may be appropriate to make explicit the existence of attached detectors since properties like measurement accuracy and measurement ranges, which are detector's properties, are not listed as instrument properties.
Fig. 2 also shows that OBOE provides concepts for describing entities of interest and their measured properties. More speci cally, measurements are of properties of entities of interests. These measured properties are called oboe:Characteristics. These oboe terms are listed below along with their original de nitions. { oboe:Entity \denotes a concrete or conceptual object that has been observed (e.g., a tree, a community, an ecological process)." { oboe:Characteristic \represents a property of an entity that can be measured (e.g., height, length, or color)." 4.2
Imagine two data sets of \air temperature" measurements obtained from a
common weather station thermometer and using Celsius to represent measured
values. These measurements still could use di erent hardware and software
congurations or calibrations for the platform and observing agent { in this case
the weather station and the thermometer respectively, thereby making the
measurements di cult to compare or use in combination. For example, during one
use of the thermometer, it was calibrated to operate in the [0o,20o] range when
the actual temperature was in the operation range. During another use of the
thermometer, it was still calibrated to operate in the [0o,20o] range although
the actual temperature was in the [-10o,10o]. As a result of a bad calibration
decision, the thermometer ended up generating data that may be classi ed as of
low quality. OBOE is aware of the impact of context in observation data
management, which is why the ontology provides a context concept. The notion of
context in OBOE provides a start for encoding context, however it does not
include descriptions of what constitutes a context property, and more importantly,
what does not constitute a context property.
{ oboe:Measurement is an assertion that a characteristic of an entity was
measured and/or recorded. A measurement is also composed of a value, a
measurement standard, and a precision (associated with the measured value).
Measurements also encapsulate characteristics that were recorded, but that
were not necessarily measured in a physical sense. For example, the name of
a location and a taxon can be captured through measurements.
{ oboe:Standard de nes a reference for comparing or naming entities via a
measurement. A standard can be de ned intentionally (e.g., as in the case of
units) or extensionally (by listing the values of the standard, e.g., for color
this might be red, blue, yellow, etc).
{ hasneto:DataCollection de nes the technical activity of the collection of data
that is empirically observed. So far, the state of the art of semantics for
observations and measurements characterizes this activity as an Observation.
The HASNetO ontologies take the position that the concept of Observation
is a scienti c activity while most if not all existing ontologies embody the
position that describe the technical activity of data collection.
{ oboe:Observation represents an `observed entity' that is, an entity that was
observed by an observer. An observation often consists of measurements that
refer to one or more measured characteristics of the observed entity.
Provenance knowledge is an important part of contextual knowledge that is
often not fully captured in many scienti c applications. HASNetO is a major
bene ciary of all the previous work developed by the provenance community in
de ning a truly general-purpose vocabulary for provenance, which is the W3C
PROV language [
Fig. 4 shows how vstoi:Deployment and hasneto:DataCollection, which are two of the most important technical activities related to empirical data, are de ned as subclasses of prov:Activity. These two subclasses of prov:Activity have been discussed previously. Below, we brie y describe prov:Activity and its two complementary classes prov:Agent and prov:Entity.
{ prov:Activity is \how PROV entities come into existence and how their attributes change to become new entities, often making use of previously existing entities to achieve this." { prov:Agent \takes a role in an activity such that the agent can be assigned some degree of responsibility for the activity taking place. An agent can be a person, a piece of software, an inanimate object, an organization, or other entities that may be ascribed responsibility. When an agent has some responsibility for an activity, PROV says the agent was associated with the activity, where several agents may be associated with an activity and viceversa." In HASNetO terms, we see that some prov:Activity instances in support of data collection are mainly performed by humans while others are mainly performed by machines. However, it can be challenging and in fact unnecessary to classify these activities as long we can fully describe the exact involvement of each agent in each activity, including the fact that the agent is a human or a machine. { prov:Entity is de ned as \physical, digital, conceptual, or other kinds of thing." In HASNetO, prov:Entity is used to represent, for instance, samples that have been collected and that are going to be further analyzed in a lab, that is where scienti c measurements and data collections occur. Also, prov:Entity is used to specify any information that is fed into an platform, instrument or detector that change the behavior of any of these measuring devices. Finally, while an instance of prov:Entity may be an instance of oboe:Entity and vice-versa, we prefer to treat them separate considering their distinct roles in scienti c activities. 5
Ongoing research activities in support of semantic sensor networks make use of
the description of instruments and detectors (many times called just \sensors" in
the literature) to maintain complex networks of sensors, while providing
integration of the collected data. In [
The concept of Observation data is treated in the literature [
O&M [
In other initiatives to annotate scienti c data, VSTO provides a data framework for ontology based discovery of datasets across the elds of solar physics, space physics and solar-terrestrial physics from multiple repositories. 6 6.1
One strength of HASNetO comes from the fact that OBOE, PROV and VSTO
are mature community-developed ontologies. For instance, OBOE was initiated
by an NSF-funded project and has evolved through a number of sponsored
research projects. PROV is a recommended standard from W3C endorsed by
academic organizations and industry. VSTOI is a by-product of the VSTO
ontology [
The HASNetO ontology may be regarded as suited for the management
of empirical scienti c data if the nal querying and browsing capabilities of a
HASNetO-based infrastructure is regarded as useful by a community of scientists
addressing some data management challenges described in terms of use cases.
Our HASNetO ontology powers our prototype HADataC (Human-Aware Data
Collection) Framework, which is under development, and has been deployed in
support of three major research projects/organizations:
{ At the RPI Tetherless World Constellation in support of the Je erson Project
developed in collaboration between IBM, Rensselaer Polytechnic Institute
(RPI), and The FUND for Lake George [
The Human-Aware Science Ontologies (HAScO) is a family of ontologies. HAScO itself is a high-level ontology that describes scienti c activities along with supporting technical activities. Within HAScO, data collections are de ned as technical activities in support of empirical and simulated data. HASNetO is the HAScO ontology that provides a vocabulary for encoding knowledge about empirical data collection. One overarching goal (and challenge) for HAScO is to provide a vocabulary small enough that domain scientists are comfortable using it, but still rich enough for use in explaining complex relationships involved in the combined used of empirical and computational scienti c activities. 7
The Human-Aware Sensor Network Ontology (HASNetO) was established as an integrated and comprehensive vocabulary for encoding knowledge related to scienti c measurements and their derived empirical data. HASNetO aligns and resolves con icts from the integration of three community-developed and community-maintained ontologies for observation, sensing, and provenance. Contributions include the identi cation of appropriate covering ontologies, the alignment between them, the gap analysis, and the gap lling. One key gap that HASNetO lled relates to providing terms for modeling human interventions related to empirical activities. Sensor deployment and data collection are examples of such human interventions. The exact interpretation of the Observation concept from the OBOE Ontology, and its meaning in terms of data collection was clari ed with the creation of a HASNetO concept called DataCollection. This is the actual act of collecting data in the context of scienti c activities such as an OBOE Observation itself and empirical experiments. It is also worth mentioning that HASNetO clari es the use of the term \sensor" in its description of a sensing infrastructure, and when it is compared against competing e orts.
Finally, a full explanation of human interventions in measurements generating empirical data is provided by the provenance of empirical data, which is de ned as a result of combinations of activities such as VSTO Deployment and HASNetO Data Collection. Moreover, VSTO Deployment and HASNetO Data Collection are de ned as PROV Activity's specializations.
Acknowledgements. The third author is supported by CNPq - Brazil - Science Without Borders scholarship.