2015-05-12 Meeting notes

Date

12 May 2015

Attendees

Proceedings20150512 FIBO BE FCT Technical meeting.docx

20150512 FIBO BE FCT Technical meeting

David introduces the challenge of variability of meanings across regulations, versus words. That is, not every use of every word in the literature, can be expected to resolve to one common meaning that can be framed in the logic of an ontology.

Today we will look at examples of this challenge in relation to business entities. Alison Charney and a colleague at a legal firm have been helping David with legal and regulatory definitions of some commonly used words. They may join these calls when time allows.

Example word: Company. This has various definitions, not all of them corresponding to a single semantic concept. Nonetheless David feels that we should have a concept of Company somewhere in FIBO, for one or other of these meanings. Elisa notes that some of this is already dealt with in annotations in FIBO. Reference: Bank Holding Act (US) 1956.

The word "State": Question: does this also include UK home nations, other countries? The Act is specifically a US Act and refers only (explicitly) to US federal states. Whether we interpret this by extension as meaning also federal provinces / states in Brazil / Nigeria etc. and / or countries like France etc., is an open question since the Act does not cover these.

The word “Sole Proprietorship”: this is given as a Company in the US act. other sources also include Sole Proprietorship under "Company". Reg W states that a person cannot be an Affiliate and Company is referenced all over the place, presumably as a thing that explicitly excludes persons. This leaves open the question of whether it includes Sole Proprietorship. David could not find an answer to this in those US regulations. Has passed this question on to other legal experts.

Elisa has spoken to a banker, who says that these things (in the holding Bank) refers only to things which own a Bank, and so the definitions given here do not require to these kinds of entities more generally. For this reason, there will be things called Companies across the world which do not comply with this definition.

FBC has a "Bank Holding Company" concept. FBC also has another similar concept of Special Holding Company more generally this remains to be debugged globally.

Agree that Bank Holding Company is a Relative Thing. However, to David, the word "Company" still represents an important distinction from other concepts we have already defined in FIBO under different names.

We have established that a Sole Proprietorship cannot be an Affiliate. An individual cannot be an Affiliate. But, a Sole Proprietorship comes under the umbrella definition of a Company. Conclusion is that there are many different statutes, each of which has its own specific definition of a concept which has the label of "Company". These are not all the same. The challenge for FIBO is how we are to deal with concepts that have this kind of variability.

The folks at Wells who are involved in international banking are very interested in FIBO being able to represent jurisdictional difference among concepts that they need to come to grips with for their global banking activities. Mike notes: many of these challenges are terminological and will require a principled application of terminological theory and tooling, On top of the simpler matter of ontology. See Straw Man on Slide 4.

Comment (Rosario) important to recognize that a company is not necessarily always the same thing a British company is one whose registered address is in the UK. The company's behavior we are only interested in certain behaviors in respect to ownership, we do not need to capture all aspects of behavior in relation to a company, e.g. who they can hire, what privacy laws they abide by, i.e. all the different concerns of the regulations here.

Rosario is saying 2 things: the Jurisdiction of a company is an integral part of its definition (regardless of where it operates) so we should not go top down on the jurisdiction.

There are 3 reference relationships among companies: ownership, debtor obligation and partnership. In respect to these 3 relations, are things we need to capture in order to validate, enforce or offer more insight on these things. So, we have to think of what can be say about the jurisdiction and regulation. That is, we should not go top down from jurisdiction. Alison Charney (one of the lawyers) made the same observation to David yesterday.

David: FIBO has been both top down and bottom up. What should drive us is specific use cases. Reg W is a specific business use case. Kinds of Affiliate were based on ownership and control relationships. Beyond that we have not got into specific regulations.

Comment (Jeff): In addition to regulatory aspects, and jurisdictional impact on behavior, also because regulatory behavior is defined by a jurisdiction, the low level regulation that is responsible for bring an entity into existence if important. For instance in the LEI, in the code list, need to define code lists for business registries. This is not trivial because as well as sovereign states you have federal nations. e.g. Germany, the US and so on, where there is both a federal and a state jurisdiction.

The concept of Jurisdiction is a specific requirement which brings a lot of material germane to defining the legal entities themselves. Mike notes that jurisdiction was particularly important back when FIBO BE had a concept of

legal personhood, since it is statute laws which frame the existence of legal persons that are not humans.

Jeff: there are other kinds of legal person, e.g. chartered entities, and this adds to the complications of what kinds of things are entities, and how these relate to the jurisdictions. Elisa: FBC covers some of these, e.g. registrars and regulatory authorities. Need to relate those to the identifiers that are registered. FBC are working through these. Conclusion that in FBC we have not captured this stuff appropriately yet.

Jeff: when it comes to the jurisdictional agencies that allow certain kinds of entities e.g. trusts to exist, there are relationships amongst things from the sovereign down, so there are relationships within the jurisdiction tree itself.

This has not been addressed in FBC. LEI are struggling with these issues, to be able to distinguish things like funds, from kinds of organization.

Slide 4: Jurisdictions influence a variance in the types and attributes of these business entities. There is some commonality between the US and the UK, but also some unique differentiating attributes for corporations, trusts etc.

Rosario UcedaSosa: But 'jurisdictions' are a different 'animal' than corporations/partnerships/etc. I don't understand the directory structure. It's not clear if this is a packages structure, a subsumption hierarchy, or some other organizing principle.

Answer: this is a directory structure. Each box represents a directory, with the RDF files at the leaf level. As well as being a directory structure, this is also a subsumption hierarchy. That is unlike FIBO generally, the structure here represents both a containment structure and a subsumption hierarchy.

Pete: does this mean we would only have jurisdiction specific RDF files and not have generic RDF files at the jurisdictionneutral higher level? No, you would have RDF not just at the leaf level as stated above, but also generic ones. e.g. a generic Corporations rdf file which would be independent of the jurisdiction specific ones. Then the jurisdiction specific differences would be lower down in the tree. This is why the proposed containment hierarchy also reflects a subsumption hierarchy. More general concepts would be higher up the tree. This is less clearcut on the Regulations part of the tree on Slide 4. These would presumably be concepts common to all regulations regardless of jurisdictions. This reflects that there is some dependency on the definitions of ownership and control, per regulations. Then for instance the definition of majority control in Reg W would be very different to the definition in other regulations, e.g. different regulations set thresholds of e.g. 25%, 15% and so on, as deemed ownership or deemed control under that regulation. We worry about the concept first, not the label or the tag.

Pete is concerned about this deep nesting structure by enforcing a monohierarchical structure, you get the same issues you get when trying to file emails whereby some things fit legitimately under more than one part of the structure. Hence while a containment hierarchy is necessarily monohierarchical, ontologies and things in the world are polyhierarchical.

David intends to come up with a precedent which can be applied across the whole of FIBO. This is the intent of the extra bit on the right of the slide 4. We can then use About files. This slide summarizes one approach that Dean and David talks about off line, for us to discuss.

Pete: you only need the right hand side, then you don't need the deeply nested structure on the left side. David sees this as incremental you could create this, then consider what domain it is associated with, do both, then (aside, would want to see a UK Corporation in the directory tree of Corporations, rather than following along this line and seeing everything about the UK in a separate domain / end of aside) so there are pros and cons to both.

This will be complicated for managing updates. If we got into multiple jurisdictions and multiple regulations it would be hard to maintain. Also don't forget that OWL does not have a concept of directories what we have in FIBO are URI fragments which we call Modules.

Other comments: How do we conceptually define something in one domain, something like a faceted model at an ontology level? How do we manage this?

Answer: preferred way is through annotations. If you create a tree structure for facets, you can do this through annotation properties and even sub annotation. Here we are really talking about a separate referential level, which is why metadata is more appropriate (Rosario). Rosario has examples of something similar that they did in Smart Cities. What they did was use annotations for 2 things: 1. to map to standards so you could get a cut of an ontology relevant to one given standard; (annotations can be both in classes and properties anything not needed for reasoning was left in the annotations). The requirement is to keep these concerns separate so that they don't contaminate each other.

Action: Rosario will provide some slides on how Smart Cities dealt with faceted ontologies.

David: looking at Protege, to show how we dealt with Reg W Affiliate, as an answer to this point. When he turns on namespace prefixes, you see specific modules. These are in a MSDOS Directory hierarchy on David's machine.

These follow the middle part of the Slide 4 structure. Realizes that regulations are probably jurisdictional. So the directory structure was tweaked to match this. So Reg W comes under Control, since the notions of Reg W Affiliate are to do with Control.

Q: Do you need to say is Reg W Affiliate: if this is already under Reg W in the urI? not necessarily the namespace would make it clear that the relationship is specific to the regulation. However, when you view Protege with the namespaces turned off, this would not be very clear. So that is why Reg W was included in the label as well.

Mike notes that isRegWAffiliateOf should not be a direct sub property of isAffiliateOf, since the Reg W relationship includes the giving of financial advice. Mike did an ontology for this, but the Reg W folks did not use it. It has unions of subproperties of the more general concepts like Affiliate and like giving financial advice, and for only the sub set which are themselves financial institutions.

David notes that the relationship isAffiliateOf itself was not one we were able to converge on a single meaning for.

Mike Atkin proposed something on this group which everyone agreed with at the time, which was that Affiliate is a basic union of the most generic ownership and being owned relations. All agreed to that at the time. Mike B's ontology of Reg W affiliate then used this to show how Reg W affiliate was a union of some subproperty of this along with relationship in which financial advice was given.

Pete: We should focus on the concepts. We should then give the concepts artificial sounding names which describe the concept as modeled, not labels that try to match the names for things in the regulations.

David asks are there examples of this as good practice in ontology?

Mike: yes. In FIBO, But people keep rejecting ontologies that are framed this way because they either see words they don't recognize or don't see words they do recognize. Then they break the ontology.

David proposes that we do not have a concept of generic isAffiliateOf because there is no concept that covers all examples of this.

Mike disagrees there is a well understood term of affiliate, just that Reg W affiliate is not a simple sub property of that. There is no reason that would problematic.

Pete has concerns that trying to do too much with the directory structure is a dead end. There is no semantic to a directory structure in OWL. For example would we build Corporation in a Corporation directory tree,

Mike notes we should not be doing terminological work in an ontology.

Action: Rosario will do a presentation on how to use annotations.

Action: Pete will come back with proposals of how to put all this in one module, including with an example. Pete's believes David already has that with the Reg W box on the right hand side of Slide 4. Pete's proposal is not to go too deep in the hierarchy and not get hung up on directories

Action: Mike to come back with an example of something that is an ontology that doesn't try to get distracted with terminological considerations.

Question (Jeff): given that regulations are often quite specific to jurisdictions (by definition),

e.g. the Napoleonic Code is one way to approach this, would be to have one very general kind of regulation very generally, and then an actor, and then a jurisdiction, and then the applicability of the rule is the association of the jurisdiction, the rule, and the person. Would this provide the means to cover a wide variety of global types of rules (like anti money laundering)? OR would that get bogged down by specific regulations in specific jurisdictions?

Yes, there may be some concept that varies by jurisdiction and by regulation, so within that box you have all the permutations of those, and then something very unique that represents the intersection of those dimensions. But, we need to figure out structurally how you would do this. How in practice would people want to bring these together for a particular purpose? e.g. all UK, just at the Reg W level and so on?

What are the use cases for pulling these parts together? David that's the question we are not sure about yet. So for instance, we would collect everything that is distributed across FIBO for dealing with Reg W. Reg W cuts across entity types, ownership and control concepts, transactions, accounting (and also financial advisory services). We have not yet built out much on accounting (see FND roadmap). So the thing on Slide 4 would then gather al these together and put them all in one module. So we would have ownership and control relations, accounting and entities, all in one module. We need to lay out all the options and see what makes sense.

If we just consider Reg W this might be quite reasonable, but it would be good to see a selection of different regulations, to see if Reg W is representative of being as complex as it's possible to get, or if it's a very simple example? Wes Moore may be able to help. We want to focus on issues pertaining to concepts in BE. Some of these relationships may go outside of BE into Securities, FBC and so on (and FND for accounting). So these would be things that cut across the other vertical domains.

David asks everyone to think about this problem and how we should solve it.

Another alternative would be if there is no About ontology. A third option would be one module for the regulation, that would import all the ontologies needed from the other areas, and have the specifics in it. Maybe there is a 4th or 5th possibility please think about this and come up with recommendations. Also please prototype those suggestions if possible. Then we can share these ideas with the different FCTs, and come up with a conclusion on how to do this across FIBO as a standard.

Action items

Former user (Deleted) Action: Rosario will provide some slides on how Smart Cities dealt with faceted ontologies.
Former user (Deleted) Rosario will do a presentation on how to use annotations.
Mike Bennett come back with an example of something that is an ontology that doesn't try to get distracted with terminological considerations.
Pete Rivett Pete will come back with proposals of how to put all this in one module, including with an example. Pete's believes David already has that with the Reg W box on the right hand side of Slide 4. Pete's proposal is not to go too deep in the hierarchy and not get hung up on directories