Date

30 Jan 2018

Attendees

Agenda

1) Use Case reminder

2) Where we are on our road map.

3) Open Action Items

4) JIRA Issues Review - https://jira.edmcouncil.org/projects/FCP/issues/FCP-4?filter=allopenissues

5) Todays content discussion.

SMIF OWL-UML

SKOS

RDF/S

6) For next week.

Proceedings:

EK blocked by GitHub issues so her work is only on local machine. New updates will be next week.

20180130 FIBO CFTC

EK We need to go through the detail that Pete has presented, but not yet in DER. Pete Item 2 shows there are 2 redundant properties. hasUnderlyingDerivitives and some others. DN hasUnderlyer is the latest DER JIRA issue.

Pete: The incoming data is a call option. Max: Should just use the word option and it has an underlyer which enables infer of if it is a put or a call.

DN Let's focus on IRS for this meeting, table this discussion until we have all correct for CFTC. E.G. Don't have definitions that we need. PR PoC is also short term and perhaps should be considered at the same time. DN Ok, but let's focus on what is most relevant to IRS. EK Let's deal with the upstream first.

DN to PR, some redundancies should be brought up to the FLT to resolve. Else, can't make use of other parts of Provisional FIBO.

DN to JN What is on your plate to discuss? Sent a draft .ppt for presentation. Wants DN and DW to look a that.

Jeff to Max, I think knowing if an option is a put or a call is necessary, since it determines the direction of the transaction between counterparties (money <==> asset ) if exercised.

DN to DA Progress since Friday? No working on M&T bank. No progress on this. Now will look at DN submissions from DN. Redundant or duplicated labels. Types of borrower, e.g. What do to? DN This is a design principle that should be discussed with FLT. E.G. Label, with different meanings creates confusion. DA There is a policy to not do this. Queried for this and got a couple of pages. Some are genuine issues. Will count for DN. Max Don't need to define borrower outside of debtInstrtument.

DA There are 42 of them to resolve. This could lead to 42 JIRA tickets. EK Mike U did not reuse work that had been done. This could be much of the problem. MB and EK need to work with MU to resolve. DA some are in LCC where LCC is colliding with FIBO.

ACTION: DA to send list to EK and others.

EK Collisions with LCC need to be resolved. Decision on what to integrate with FND. Because LCC was outside of FIBO some labels had to be recreated. DN Need to look at the list and prioritize.

ACTION Create a JIRA ticket to do the resolution. PR already made some suggestions.

DN We had agreed to prototype some of this already. We need more uniqueness in range because else we are not able to align a source element from a relational DB that is quite specific, and we try to align it with a generic. The predicate is not enough to create a 1:1 alignment. A primitive like date can be tied to hundreds and hundreds of specific columns. We are taking some labels built into predicate and auto generating a new object. DN Want to be able to look at a particular semantic element. E.G where are all the instances of date? Need to be able to narrow down to a particular date. Want to be able to query to instances in the graph. Look for concepts that are dates of issuance. Need to tie a column or a set of columns to a semantic object and see the fan out of relational columns.

Pete and Jeff Use the predicate and you can get the same thing.Pete There is no mismatch. The predicate in FIBO matches the column in the relational database.

Jeff Braswell: There are physical data types ( Integer, VarChar, Boolean, Date, etc.) but those are fundamental to CPU processing, and are not the same thing as the logical elements for the physical bits.

Max The impedance mismatch is that need to traverse the graph to find that. DoB is related to a person. Jeff Braswell: But you would map the 5 different labels for 'issuance date' to the 'issuance date' property in FIBO, not all things that are of type date, no ?

DN Each SDR has a set of columns that are tagged differently. We need to match each of the 5 different columns that are really the same thing. Need to match date to each of these 5 elements. May have other dates coming from the SDRs.

DA R2RML basic mapping does what DN wants.

Jeff Braswell: Other physical date types would have different logical/semantic properties, not just generic 'date'. DA Anzo does or does not use R2RML. Take a column and map to a predicate and then look at the type of the object to see where to look next. No need to multiplex on the types.

DN shows an example. of what he wants to do.

DA Take the predicate that it points to and do the R2RML direct mapping. All is automatic.

DN We are not linking to Predicates. We are linking to Objects. DA Yes we are.

DN R2RML shows how elements are tied together, not have seen what I need.. DA Seems that DN is talking about other tools that are not semantically aware. DN talking to a vendor tomorrow. Will see if they have a way of linking by predicate. DA in the sense that we can generate what DN wants, I can see this. Not redesign FIBO.

JB is DN talking about Global IDs. Yes. JB, they are not a semantic mapping tool and have a unique issue. Need to start with a model that has semantic content. Global IDs way of handling associations is not scalable. They will always have issues that require people. DN the fact that Global IDs does not connect to a triple causes a problem and we will have certain primitives that must generate objects.

PR is CFTC constrained by a vendor. JN No.

DN Need to work though some issues B Think that JN does not need a data profiling tool. DN agrees.

Decisions:

Action items

Dean Allemang to send list of the 42 duplicate labels to EK and others.
Dean Allemang Create a JIRA ticket to do resolution