INTERVIEW
Toby Considine and Therese Sullivan

Toby Considine, TC9, Inc
Therese Sullivan, BuildingContext Ltd

Semantic Spaces: Where Everything is in a Name

An OBIX semantic space accommodates naming 'things' with tags from multiple semantic libraries.  The goal is not a universal fundamental ontology (UFO), but rather a taxonomy that allows the lexicons of different specialties to coexist and support one another, while keeping the byte-size of any name small.

Last month, Toby Considine closed his column with an invitation to meet him in Northern California in mid-September, where he was speaking at the TechIntersection conference. I seized that opportunity to interview Toby on the topic of  Semantic Spaces and the Buildings IoT. Some of Toby’s authority on this subject stems from his work since 2006 as co-chair of the OASIS OBIX Technical Committee. OBIX is an unencumbered web service designed to interface between building systems and enterprise applications. Toby is also active in many allied efforts, including the National Building Information Standard (NBIMS) for the design and construction of buildings and in national efforts to define the Smart Grid. He was also co-champion of the FIATECH Information Technology Roadmap Element 5 (“The Intelligent Self Maintaining, Self Repairing Facility”). It was a great conversation, the key parts which I’ve transcribed for you here:

Sullivan:  What are OBIX tags? How can those in building automation and control use them for open building information exchange.

Considine:  Naming is the essence of communication. You could say that all knowledge derives from over-laying of names of different semantic spaces.  In OBIX, we created a semantic space for the inclusion of multiple semantic libraries. Using OBIX means that whenever you create a tag for data, you identify the library it came from by an extension. For example, a ‘thing’ from the HVAC realm like an air handler or sensor could be described with a name from the Haystack library and, thus, could be given an ‘h’ extension.  The whole tag can be small - because OBIX likes data to be small.  Once you identify a semantic library as being within your semantic space, you can tag any ‘thing’ with that extension. 
There are many semantic libraries. And there is no limit to the number of tags a ‘thing’ might have. Even something like an air handler might have seven tags from three different semantic libraries. There are cases when, in addition to haystack tagging, you would want to add OmniClass tagging, Department of Defense tagging, medical equipment tagging, etc.

Sullivan:  Why would you ever want tags outside the Haystack system for an air handler?

Considine:  Here’s an example of why. A few years back in a big metropolitan hospital in Cleveland, an energy engineer was hired to optimize the way the blowers on the doors of the main entry were operating during cold weather. The purpose of the blowers was to warm and dry people as they entered the facility from the cold, wet, snowy weather outside. There was an unforeseen consequence of the newly optimized structure. All of the sudden, through the lobby, down the stairs, through a tunnel, through another building, up an elevator shaft, and into the hospital’s Burn Center, there was no longer positive pressure.  In a Burn Center you are an open wound over a large part of your body. It’s critical that air not enter the space. So the mechanical engineer specifies higher air pressure in the Burn Center, such that when someone opens the door the air flows out. However, in this case, because of pressure changes a couple of buildings away, this was no longer the case. Burn victims died before the problem was understood and reversed. 

Haystack tagging would tell you what the pressure is locally and that the front-door has a system blowing. It might give you a lot of clues as to how the systems are working. But knowing that the blowers are affecting an area that requires positive pressure is a medical classification. And I don’t think that you should put the medical space classification into Haystack.  Yet, the equipment that serves the area should be classified with appropriate medical-space tagging. While this is a dramatic example, the need for other types of tagging related to the purpose of the building space is not unusual. Factors like temperature, humidity and pressure of air in manufacturing areas can also play a key role. So the semantics of those with specialized knowledge of the manufacturing processes are typically the best source of appropriate tags for equipment used in those areas.

Sullivan:  OK. But, just when naming all the equipment and devices that interact in a basic building, wouldn’t it be better to have one foundational library that includes all tags? Shouldn’t we be driving toward a single semantic library?

Considine:  In ontology circles, we have a joke: “Universal fundamental ontologies (UFOs) are always being reported; but, they are always blurry in the pictures.”  The reason is they don’t work.

A few years back, I was working with a group that had labored for years on their information model. We had determined that it would be useful to exchange weather information, predictions and observations, with outside parties, information that could be used to plan smart building operation and understand prior results. The information model already a small nugget of weather information, temperature and humidity if I recall. The sensible approach was to turn to an organization like NOAA to contribute weather semantics. One participant suggested that NOAA be “made” to use his committee’s information model. Fortunately, we decided to rely on the weather experts, and they are still working on a common information format that can be used by smart buildings, smart grids, smart transportation, and anyone else who might want their IoT application to understand weather.

[an error occurred while processing this directive]In the smart buildings realm, Haystack serves the HVAC community in a nimble way. Smart grid and energy services have invested in their dictionaries. US BIM (building information modeling) standards originated in the tri-services: Army, Navy, AirForce. There is now an international standard for naming layers in CAD. In international BIM there is a common framework for describing components of a BIM of interest to operations: the Construction Operations Building Information Exchange. COBie identifies spaces by their use, Equipment and Systems by their type, and links them together by Zones, which are collections of spaces with a common purpose or linked to a common system. But we have elevators and England has Lifts, and the French first floor is the first one you take stairs up to. The US variant of the standard specifies OmniClass as the source for names for equipment and purposes. OmniCLass is particularly useful because it provides a taxonomy, that is, a hierarchy of equipment types and space purposes.

You do want to have standard names; but where does the standard name come from?  That’s the challenge that multiple semantic libraries deals with.  This is why in OBIX you can include an extension to any name-tag that tells you what library or dictionary it comes from. OBIX does not say “You cannot speak French anymore.” OBIX merely requires that you tell us when you are doing it.

The tension between the inconsistent views of different semantic sets is where ontologies live. There are legitimately different ways to look at [a building]. If you remove all the tensions, you remove all the wisdom that ontology implies. Think of names as 3D objects. If you squash them down to 2D, you lose something.