Testbed-12 General Feature Model Engineering Report

Given that almost all data has a spatial-temporal component, this OGC® Engineering Report evaluates the degree to which the General Feature Model can accommodate complex analysis on large multi-disciplinary, heterogeneous data collections for the purpose of extracting new information from dissimilar content. It looks into the idea of defining a Spatial-Temporal Service and Analytics platform and identifies the data types, operations and architecture patterns would be necessary to spatial-temporal enable any type of content.

This OGC® document reviews the General Feature Model and gives guidelines for necessary modifications to broaden its scope, so that it can be re-used for non-geospatial centric applications and extended as necessary into a general model for all object types.

The idea to integrate information and capture knowledge has evolved in OGC over a long period of time, accommodating the continuous changes in technology and the range of services to be integrated. The following OGC documents summarize some of this work and have been considered in the development of the initial concepts:

The Web Object Service Implementation Specification (OGC 02-049 / 03-013) aimed to satisfy a requirement of the OWS 1.2 project to manage many different types of objects, including styles, symbols and images, using a repository interface. “The specification defines a set of base XML types that define the behavior of a Web Object Service. A Web Object Service is a generic web-based repository interface. The interface supports the following operations: GetCapabilities, DescribeObjectType, GetObjectById, GetObject, LockObject and Transaction. The specification assumes that the distributed computing platform is HTTP and may define both XML (suitable for the POST method) and Keyword-Value Pair (suitable for the GET method) encodings of each operation.”

The OGC® Testbed-10 Service Integration Engineering Report (OGC 14-13r1) “specifies a means of discovering and describing associations between web resources (both OGC and non-OGC). The discovery of associations is accomplished by means a new operation named GetAssociations. The description or representation of associations is accomplished by the definition of an XML Schema for expressing associations in XML. The scope of this document is limited to the existing OGC web service framework, which predominantly uses XML. However, it is anticipated that over time the association schemas defined in this document will be extended to provide JSON snd JSON-LD expressions which are rapidly emerging encodings within and outside the OGC.”

The OGC® Testbed-11 Incorporating Social Media in Emergency Response Engineering Report (OGC 15-057) studies two different approaches to incorporating social media information into Emergency Response applications: “The first [approach] was based on the concept of using an OGC Sensor Observation Service (SOS) for handling humans as sensors is illustrated from an interoperability perspective. The developed O&M-based data model is presented and illustrated with examples from different social media platforms. This is complemented by a description of how data loading from social media platforms into an SOS server can be achieved. The second approach was based on using Linked Data to integrate social objects produced by the different social media networks. The ontology for describing social objects and activities is based on SocialML ontologies that can be extended to accommodate new activities and social objects. The integration of social sites was done through the use of RDF scrapers accessible through a REST API. The resulting knowledgebase was made available through a GeoSPARQL endpoint. Another important topic of [OGC 15-057] is the use of the OGC Web Processing Service (WPS) for analyzing the available social media data (i.e. detect clusters). “

However, the requirement to integrate heterogenous data using spatial-temporal attributes has also been addressed outside the OGC. One example explicitly mentioned in the Testbed-12 RFQ is the FGDC endorsed Time Space Position Information (TSPI) Version 2.0 standard (NGA.STND.0019_2.0), which "provides a single means of encoding spatiotemporal information for the storage, manipulation, interchange, and exploitation of spatiotemporal data. It is designed as a set of reusable data components upon which both Geography Markup Language (GML)-based application schemas and non-GML-based XML schemas may be developed. It specifies a registry-based extension mechanism enabling the development and reuse of additional spatiotemporal XML Schema components. It integrates the XML Schema for the United States Thoroughfare, Landmark, and Postal Address Data Standard, FGDC-STD-016-2011, including the necessary updates to work with the OGC® GML 3.3, Extended schemas and encoding rules."

The Testbed-12 General Feature Model ER now takes a step back from specific implementation requirements and looks into the principle capabilities of the GFM to accommodate non-geospatial data, information and associations as features.

A simple use case scenario is based on observations from non-geospatial sources, subsequent decisions based on selected observations and a requirement to create associations, which allow a user to access all associated features. The objective is to make a body of evidence for a decision accessible from various angles. This can be the traditional geospatial approach of showing a decision on a map and providing a report of associated features. However it should also be capable of providing graph-like overviews of features and associations accessible from non-geospatial systems.

All questions regarding this document should be directed to the editor or the contributors:

This initiative has demonstrated the feasibility of a Spatial-Temporal Service and Analytics Platform based on the General Feature Model. However, much work remains to be done to fully define and validate that platform. The proposed next steps in this effort can be found in Section 9.

The work carried out in this activity has not resulted in any change requests towards existing specifications.

Attention is drawn to the possibility that some of the elements of this document may be the subject of patent rights. The Open Geospatial Consortium shall not be held responsible for identifying any or all such patent rights.

Recipients of this document are requested to submit, with their comments, notification of any relevant patent claims or other intellectual property rights of which they may be aware that might be infringed by any implementation of the standard set forth in this document, and to provide supporting documentation.

As part of their homeland security mission, NGA has compiled the HSIP Gold database. HSIP Gold is a compilation of 560 of the best available geospatially enabled baseline infrastructure datasets for all National & Defense Critical Infrastructure Sectors. Now they want to extend that capability by providing not just feature data, but a comprehensive integrated information environment. This environment provides a comprehensive suite of integrated information as well as analytics to derive knowledge from that information.

Central to this capability is a data aggregation service: Data is collected from open source, crowd sources, and procured from key data providers.

All data is assigned a unique identifier, and populated with discovery metadata. That metadata includes the spatial and temporal extent. Core to the data store is a global feature data set. This foundation feature data is a detailed representation of the global real-world objects. The foundation feature data provides a spatial-temporal context against which all other data can be understood.

The data holdings do not stop with data. They also include the results of analytic processes applied to the data. These results are classified as observations or judgments.

Observations identify correlations between data. These may be generated by observation or through automated analytics. Correlations are statements of fact, but do not indicate meaning.

For example, two suspected criminal operatives are in San Francisco at the same time and stayed in hotels within ½ mile from each other. This could indicate that they are planning an attack. It could be simple coincidence. The observation is a simple observance of fact. It makes no judgment as to the significance of that fact.

Judgments are the result of analytic processes, which develop conclusions from a body of evidence. They can be the result of automated or human analytic processes. A judgment must include:

This information is sufficient for a user to review the analytic process and make their own assessment as to its value.

Data, observations and judgments reside in the integrated information environment and all information is spatial-temporally enabled.

This information can be compiled for distribution through a report, i.e. a compilation of observations and judgments intended for human consumption. However, such reports do not have to be in a document format. A report may also be an interactive web site, which allows users to visualize and navigate through the judgments and reasoning. They may also be dynamically updated as the underlying information changes.

Users can build their own analytics or use the analytics provided by the service. Analytics can be assembled into workflows resulting in more complex analytics.

This concept is also well described in NGA’s presentation ""Everything you wanted to know about the NSG Application Schema (NAS)"". The NAS document describes three major components of this integrated information environment:

Based on the above mentioned conceptual ideas a feature can be almost anything. To illustrate the concept, the chosen data for the Use Case Scenario described below includes analytical observations, judgements, tweets, images and even a police radio message.

The crucial step is to move from individual features to linked features. The association describes the relationship between features and is thus the binding connection, linking objects with each other. This GFM approach works with every type of relationship.

The topic of associations has also come up in other TB12 activities and is further described in TB12 - Evaluate Semantic Enablement ER.

To evaluate the degree to which the GFM is capable of capturing information from social media and related human judgments and decisions as features, a simple use case scenario has been developed:

Apart from the fabricated incident at the junction, this is based on a Climate Change Protest March with 5,000+ participants. The event took place in Oakland on 7th February 2015. It was well covered by tweets before, during and after the event via @DontFrackCA.

The use case scenario and it’s associated data cover all three phases of a planned event:

During the pre-event phase, information can be derived about the topic, location and date, as well as the anticipated size and nature of the event. In the case of a planned event, where organizers and local authorities work together, this might provide information to judge the degree to which planned and actual developments align. In case of an ad hoc event, it could provide early warning indicators to trigger actions for preparedness.

During the actual event phase, social media information can be used to support situation awareness and handling. For TB12 we have inserted a tweet message reporting a suspicious package next to a junction the Climate Change Protest March plans to pass on the planned route. This requires organizers to make decisions how to proceed. The decisions are based on the current location of the crowd, which could be machine-processed information from tweets, human observations reported by radio and judgments on available information. Observations, judgments and decisions shall all be treated a features with a traceable relationship to each other.

Post-event information might be relevant to rate the effectiveness of decisions resulting actions.

Given that the scope of the GFM activity in TB12 is to proof the concept of storing heterogeneous observations and their associations in a GFM based model, it was decided to break down the scenario to a very simple chain of events: