2015-05-04 Meeting notes

Date

04 May 2015

Attendees

Goals

Discussion items

20150504 FIBO FBC FCT

Elisa summarizing current state on various kinds of market participant. Related to this, there were some issues in FIBO Foundations and / or BE, which caused problems when trying to reconcile the concepts with the requirement for Wizdom's LEI offering. Elisa notes there were issues with some things being or not being "thing"type in Foundations. Today, will look at some of those things, and some of Citi's requirements.

Walkthrough via Protege: Elisa will have draft ontologies next week that people can look at. Will not be a complete draft spec but will have diagrams and definitions in it, along with some OMG boilerplate stuff.

Action: Elisa will aim to have as complete a draft as possible for Berlin.

We can bring a complete or near complete spec in June for the FDTF to look at, so as to socialize it with the Finance Domain Task Force. This should 1.5 or 2 hours. Also FDTF needs to include tie for a report back on the IND FTF.

Protege demo of the ontology: At the top level not much has changed. In the Languages part, there have been recent changes. Describe not only language codes, but languages themselves. These include concepts such as artificial language, as well as natural languages, extinct languages etc. (this goes beyond what was in legacy FIBO, which defined just the simple concept of a language). Ancient languages are also included provided they have a body of written literature. The model also has treatments for dialects and the like. Also language families. And living languages, which is what most financial information is conveyed in.

These concepts cover all the classes of thing which have a 3 letter language code in

the ISO language spec. These concepts do not include programming languages. Therefore we should have a higher abstraction in FIBO for "Language" which would be the parent of all language, and rename the ISO "Language" lass to HumanLanguage. Note that living languages have an English and a French name in the ISO standard (along with presumably a name in themselves, for those other than English and French). This is the "Indigenous name" feature. We have UTF8 encodings for all the characteris for all of the indigenous languages.

Action: MB will work on the more abstract Language concept as a Foundations FCT deliverable.

Does the ISO standard identify the countries in which a given language is an official language? No, this is in the countries standard, which Elisa has also started to integrate. Country code may be integrated directly into the Countries ontology in FND. There are also official names, short names, codes, etc. for regions, islands and the like in various countries. Here, "Administrative Language" hooks into the Alpha 2 code for languages maybe also the Alpha 3 language code, we'll see. For this version we are using only the English. These are British English since that is

the working language of the ISO standard. For countries, this always or almost always the same as the US English which is the working language of FIBO. However, as a rule, for external standards we will use the language those standards are framed in i.e. British English for ISO. See historyNote for this. The plan is to submit those once we decide whether or not merge the Country representaiton.

Jeff Braswell: I believe ISO31662 (region codes) uses the 2character ISO3166 country code

Elisa: These are integrated in the Places module in a new ontology for country representation. This covers ISO 3166 and the UNFAO representations / concepts.

So UNFAO, which covers the concepts and is complimentary to the ISO stuff, has

been reinstated long similar lines to what we had in the original FIBO. We could use Country Subdivision, and with some local tweaks, make federalrelated

subdivisions (states, fed capital areas) sit alongside non federal types of subdivision.

Jeff Braswell: legal jurisdictions and 31662 regions are not isomorphic

Elisa: We would then be able to manage the relationship between these sub divisions and jurisdictions. So for example subdivisions of France all have the jurisdiction France, whereas sub-divisions of the US each has its own jurisdiction.

Relationships involving currency e.g. the Eurozone (also countries which use the SA

and so on), would be a separate, non mutually exclusive set of stuff.

Question: Why not simply use the {fr} etc. notation? We will use both.

Would we want to reify a predicate for every language? Why not use the RDF

approach? The RDF approach uses lowercase rendition of the Alpha 2 codes for the languages.

Jeff Braswell: legal jurisdictions in Germany (for registrations of legal entities, for example) consist of approximately 155 court districts e.g. {aa} refers to the Afar language. The reason is that it allows for analysis across these concepts where needed. People are not forced to use these in individual operational ontologies if they do not need to noone is forced to import all these features.

Elisa: The Countries ontology does use these Alpha 2 codes to refer to administrative languages for a country. So you have 2 options:

1. point to the string element in the code, or 2. Ignore it and tag it any way you like.

The option to construct those codes would use the 2 ISO standards (ISO 3166 and the language one), and also optionally 4646 which is the IETF thing. We can do both, or not, depending on what people consider is useful. Since this work has been done, it would make sense to use it. Use Case would determine what gets stood up in the operational ontology.

Question: when using a given external ontology, do we need all the complexity that that ontology provides, and secondly: Where a given external ontology is built on a different basis, should we stand it up as it stands or integrate with the broader FIBO concepts? Elisa: at present these external ontologies are separate (but see note above about possible tweaks we have already suggested).

So for instance the hasEnglishShortName and such properties in the external ontology, would ideally be children of FIBO hasName. So these would be standalone but linked back to the abstractions in FIBO. Then we can e.g. look at LEI requirements, which explicitly require countries defined as per ISO 3166.

Do we need all of the given external ontology (in this ase, ISO 3166) or just the bits we need? Two points: 1. does it upset the reasoner, and 2. Does it cause problems with people comprehending it?

We look at an example: ISO 3166 has Country Codes and Region Codes. The English versions of each of these is already encoded in the ontology. The English part has Alpha 2, alpha 3 numeric code, along with short name and official name (e.g. all those "Peoples Republic Of...)

Should we have a standard in FIBO about whether it does or doesn't comply with FIBO, and secondly whether or not we use the whole thing? Whether we use a whole spec or not should be taken on a case by case basis.

There are 3 treatments in the Foundations Standard as it exists. Elisa proposes that this ISO 3166 is a 4th case there was no existing ontology, just a document. So in this case, the ISO 3166 ontology is a new ontology which corresponds exactly to the written ISO document. This is kept separate and with an Equivalence relationship between Country in this ontology and Country in FIBO. That Equivalence assertion is in the new ontology, not in FIBO Foundations. The onology forISO 3166 was built by Elisa 15 years ago and has been maintained for certain other parties

Comment: If we want to shop FIBO around, we need a better story about how we

keep those things updated and how we keep consistency with the external standard that our own ontology claims to represent. There may be different ways that a given written spec could be represented in an ontology.

Would it not be better to have a FIBOspecific way to represent a given standard,

rather than have some work that does not have this visibility of how the relationship between the spec and the ontology is maintained? There is a risk whereby if we have now made this part of FIBO, we are taking on the

whole maintenance responsibility, with a need for a governance process, procedures for tracking future updates to the ISO standard and so on.

Elisa: there are two possible approaches to this. One is to retain this only in Pink and use for testing. However, Elisa notes that without reference to London you can't have LIBO (the london Interbank Offer Rate). So if we need all those assertions in order to define LIBOR, and these specifically refer to ISO 3166, then we need to refer to these. We have to be careful that when we import the standards, we should "snapshot" it (per the FND Annex on external ontologies). We also simply say that we refer to these codes, as codes, without need to make these an ontology at all. These ISO standards change slowly, and we should aim to keep them current, and so we would have to add some (new?) metadata saying that this was current on a given date.

Jeff Braswell: At the same time, from a 'rates and indices' point of view, LIBOR (and the various components of LIBOR term structure) can be identified simply by reference key. For these ones it would make sense to do that. For the Language Codes, the RA for the ISO standard, is the US Library of Congress. The LoC would like us to maintain this for them. We have reached out to the RA for Country Codes but they have not responded.

Elisa: This makes sense, but we need to document it. For instance, if there is an

external body that is the RA which has agree to maintain these, we can point to the fact they are doing it. (However, Elisa notes above that "we" are maintaining it for" them", not they for us. Note that the ISO TC (which is TC37) also has a representation at OMG.

Mike asks: who is "we", OMG or the FIBO Leadership Team? If the ontology was maintained by the standards body it would be better to have them maintain it but there are no published OWL ontologies from these bodies.

Mike notes that for purely extensional standards, there is no conceptual modeling

component to us maintaining an "ontology" since it's a simple transformation of extensional lists from one set of tables, to a OWL rendition of the same extensional stuff. Conversely, with our intentional standard (like the language one) we would have to be maintaining a truly semantic (intentional) model so then we would be dealing with an entirely separate challenge. Wherever we decide we will maintain an ontology which we commit to maintaining in synch, we would publish a statement of how we do this, and we would likely update it every year. Things with different change cycles should be put in different places. These are not in the same place, but in a separate ontology.

It might make sense to put these in a different module, so that the maintenance cycle

can be maintained separately for separate modules, (still within the same Domain!). Meanwhile individuals are maintained in separate ontologies anyway. The codes for the countries and the languages are all individuals. Also the Alpha 2 and Alpha 3 codes are kept separate and the subdivision codes are kept separate. These are partitioned along the boundaries of the ISO specifications.

Questions: are the conventions we are talking about such as having individuals in a

separate ontology has that been documented in FIBO Foundations? We're not sure, but it should be, also Dean's "Patterns" thing, which is in one of the wikis, has not yet made it into the written FIBO Specification.

Action: Elisa will find the Dean thing about patterns and post the link.

How do we manage this at present? Elisa looks for patterns in the OWL. Mike looks for conformance with the archetypical ontology design patterns. There is a lot of our methodology that is in the Methodology document, some of which it might be appropriate to add to the Foundations written spec.

Testing what do we test for e.g. do we test against patterns? We test for reasoning errors and pellet antipatterns called lints. We are starting to consider how to ensure conformance to the conceptual archetypes (ontology design patterns as generally understood).

One last review point: New material that adds a business registry, entry in the registration entry, other stuff as needed for LEI. For many institutions there is not one specific address. In some cases, there are multiple registered addresses. (Mike notes this is actually incorrect the word register would be used in many ways for many laws, but legally incorporated entities have one registered address, which is the one under which it is registered under company law, and there may only be one such. There is a physical address format that has been defined as a Business Registration address or something. This also refers to the ISO country stuff.

In ISO 17442, the specification does not specify the format but it does specify that you have to specify the legally registered address, as well as the HQ address. Therefore Jeff notes it would be misleading to use Registered in this more general

name, as it would cause confusion.

Many addresses do not confirm to the ISO 3166 code, since the legal bodies in the

various jurisdictions under which an entity is formally registered, (the Registered Address) , cannot be expected to always conform with the LEI standard for these addresses.

So Registered Address is very jurisdiction specific, and the ISO 17442 address is

specific to the LEI standard. So we can't conflate these. The LEI data file format is an interim format and likely to evolve it represents a very 'lite' approach to standardizing these. And of course the concept of Legal Entity in the LEI sense does not exactly correspond to the FIBO concept. Need to find a good way of managing this.

Action: Team need to create test individuals for these.

We meet again in one week (11th).

Original agenda for that was review draft FBC. We would not be able to review all of this but can expect to review a draft of at least the body. Expect something by the very end of this week. So that is still on the agenda for 11th. Participants are welcome to join the Thursday call for Loans as well please mail Richard or Elisa for details

Proceedings

Meeting Presentation – FBCUpdate-20150506-Part1.pptx. FBCUpdate-20150506-Part2.pptx

Action items

Elisa Kendall Elisa will aim to have as complete a draft as possible for Berlin.
Elisa Kendall: Elisa will find the Dean thing about patterns and post the link.