Metadata glossary team reviewed definition and explanatory text of GSIM (v1.2) information objects and made following comments for next GSIM revision team to consider.

Some writing convention applied to definition include:  

  • Spelling : UK English will be used (example: organisation instead of organization)
  • No leading articles in definitions : All starting "A", "An", "The", etc. will be removed. Definition of term should be able to substitute grammatically; starting with article/using sentence will not allow this
  • Definitions will not start with the concept to be defined (e.g. "A Classification Family is...")
  • Definitions will start with lowercase and have no ending dot


Base Group

  • Agent: (Meeting 16 July 2019) 1) we found "some form of responsibility" is vague, we understand that it was borrowed from prov ontology (during GSIM revision). Can this be made clearer in explanatory text?; 2) can Agent be non-Individual or non-Organisation, e.g. programme in automatized process (in this case, "some form of responsibility" does not make sense)?; 3) consider adding reference to prov ontology as source of definition in explanatory text (see how it is done for Agent In Role)
  • Change Event: (Meeting 16 July 2019) the way definition is written does not conform with writing convention. We suggest to change definition to "indication of the occurrence of a change to an Identifiable Artefact". Second sentence ("It relates to the information object(s) that has(have) been affected.") to be moved to the explanatory text.
  • Change Event Tuple: (Meeting 6 August 2019) The second sentence should be moved to the explanatory text
  • Identifiable Artefect: (Meeting 6 August 2019) There is a similar definition of this concept in the SDMX Glossary. What should be our philosophy? Keep the two glossaries totally distinct or should we refer here to the SDMX Glossary (e.g. in the explanatory text) ? We could also adopt the SDMX definition and push the current definition to the "explanatory text" section; Create GSIM Issue Need definition & explanatory text of 'artefact' to clarify the definition (& provide reference to SDMX definition)
  • Maintainer: (Meeting 6 August 2019) Should we mention here the concept of "Maintenance Agency" from the SDMX Glossary?; to move example (e.g. …) in the definition to the explanatory text
  • Owner: (Meeting 6 August 2019) Proposal: second sentence to be moved to "Explanatory text"; to add 'authority' to include working groups, task forces etc. (i.e. instead of "A statistical office or other authority ..", "statistical office, authority or other organisation...")
  • Role: (Meeting 6 August 2019) to change responsible function to responsibility function as functions can't be responsible


  • No labels

5 Comments

  1. Jenny Linnerud

    20 Sep, 2019

    In Statistics Norway we have addded System as a subtype of Agent. This is in line with the Class: prov:SoftwareAgent in prov ontology.

  2. InKyung Choi

    22 Sep, 2020

    (GSIM task team meeting, 16 September, 2020)

    Team discussed on Agent, Role and Agent In Role. Discussion and conclusion include: 

    • Agent can be a system. For example, in automised processes, we set up everything and process happens without any manual or human intervention. In this case, the system is performing a process (e.g. deriving some variable) and acting instead of people (i.e. Individual, Organisation)
    • => Action: to add "software system" as example of Agent in the explanatory text. 
    • Term "responsibility" in definition of Agent ("someone or something that bears some form of responsibility") is vague and can be confusing when we include software as Agent. It is hard to imagine a software system can "be responsible". From organisational or legal point of view, this cannot be certainly true. People reading the definition might interpret this way. GSIM should be about more broad sense of responsibility, as something that sets others in motion.
    • PROV has "Activity was associated with Agent and Entity was attributed to Agent". But saying "someone or something that is associated with an entity" is very vague. 
    • Role is the responsibility that we are attributing Agent for, Agent plays Role and the responsibility is the Role that they play. For Role, we only have very few examples of  common Role (e.g. Maintainer) which do not cover situation for software system (e.g. they can play role of updating, creating, deleting). 
    • In context of software, responsibility is the function to start certain activity, we can say that as explanatory text.
    • => Action: to add "software invocation" as examples in the explanatory text of Role
    • => Action: to replace "bear some form of responsibilities" with "plays an active Role" in the definition of Agent
    • => Action: to change "The responsible function" to "responsibility" in the definition of Role
    • Shouldn't we include different levels of process other than just saying "Business Process" in the definition of Role (the same for Agent)? While GSIM Business Process can be used for many levels, when people read this definition, they often associate it with GSBPM business process, which has very wide scope.
    • => Action: to change "Business Process" to "Business Process (or a part of it)" in the definition of both Agent and Role
    • Is it possible to use "event" instead of "Business Process"? → This ties all Agent does with Change Event. In fact, current formulation where Agent In Role requires Change Event seems to be wrong. This is too strong requirement, we should be flexible enough to allow something to happen without being captured as Change Event (not everything has to be recorded as GSIM object). Every time Change Event happens, it has to be by Agent In Role, but not the other way around.
    • => Action: to change cardinality (Change Event → Agent In Role) from 1 to 0,1
    • In explanatory text, PROV-Ontology definition for Agent  is included

    The updated definition and explanatory text are: 

    Object

    Definition

    Explanatory text

    Agent

    someone or something that plays an active role in a Business Process (or a part of it), in the existence of an entity, or another agent's Business Process

    An Agent may be an Organisation, an Individual or a software system. An Organisation may be an entire organisation or entities within a larger organisation, such as departments or divisions. An Organisation may have sub Agents, which may be either other Organisations within the parent Organisation or Individuals that belong to that Organisation. In the PROV-Ontology from W3C, “Agent” is defined as “something that bears some form of responsibility for an activity taking place, for the existence of an entity, or for another agent's activity”.

    Agent In Role

    Agent acting in a specific Role

    In the Organisation Ontology from W3C Agent In Role is called a “Post”.

    Role

    responsibility for a statistical Business Process (or a part of it) or an entity

    Examples: Owner, Maintainer, Contact, software invocation.

  3. InKyung Choi

    06 Jan, 2022

    (GSIM task team meeting, 4 November, 2020)

    Comments from Italy about Base Group

    • Agent In Role: Taking into account the communication paper the agent in role represents the relationship between the three elements agent, role and identifiable artefact 
    • Event Change: In our opinion it doesn’t add useful information to understand the meaning of Change Event
    • Contact: Contact is subclass of Role so cannot be defined as a person
    • Maintainer: Maintainer is subclass of Role so cannot be defined as a person
    • Owner: Owner is subclass of Role so cannot be defined as a statistical office or authorithy

    Discussion on the comments

    • Regarding Agent In Role (comment to add ".. with respect to a specific Identifiable Artefact" to definition): it is true that Identifiable Artefact is administered by Agent In Role, but there can be multiple (not a single IA) as indicated by cardinality (1..*). Also, it seems the addition is not so necessary. Check with Laura where Agent In Role was mentioned in GSIM Communication Paper
    • Regarding Change Event (proposal to delete "Change Event Tuple is used to list the IAs that are the source of the change and the IAs that result from that change"): It is true that the description is about Change Event Tuple and not necessary for explanatory text of Change Event => decision: agree to delete the text; also change description of attribute "Change Type" from "The type of change that occurred" to "The type of change that occurred during the Change Event" to be consistent with other attributes 
    • Regarding Contact (proposal to change definition from "person(s) responsible for.." to "type of Role in which person(s) is(are) responsible for.." (similar comment was made for Maintainer and Owner too)
      • It is true that Contact is a sub-type of Role, hence it cannot be person, but we should remove "type of", it is a Role already, we don't need it. Also, we should replace "persons" with Individual which is GSIM object
      • Can we replace "persons" with Agent to make the definition more general and open to possibility to have non-personal entity assuming the role (e.g. API playing Role of Information Provider)? => Agent has two sub-types (Individual and Organisation), we could have non-personal Agents, but we want to limit who can assume the role for certain sub-types of Role, such as Contact
      • When making changes, make sure to do it consistently for Information Provider and Information Consumer too

  4. InKyung Choi

    04 Dec, 2020

    (GSIM task team meeting, 25 November, 2020)

    Discussion on Base Group 

    Maintainer

    • Certain roles, such as Maintainer, are always performed by people (not software), hence it is better not to use Agent in the definition as Agent can be interpreted as software. => do not replace "persons" with Agent
    • Consider RASCI matrix (Responsible, Accountable, Support, Consulted, Informed) when reviewing definitions (in GSIM, we often use "responsible")

    Identifiable Artefact (IA)

    • SDMX (v2.1) glossary defines IA as "Construct that contains structures capable of providing identity to an object". "providing identity to object" is essentially what GSIM definition says but GSIM definition clarifies what structure means (in SDMX, "structure" is not defined separately) => Keep IA definition as it is, we could add in the explanatory text that they are similar. 
    • Regarding "abstract class" in the definition: although in GSIM, "object" is primarily used (not "class"), using "object" in this definition could be misleading because IA is not going to be instantiated and "object" are usually run-time instantiation of something

    Administrative Details

    • There are two attributes without description: "Release date" and "Termination date". "Release date" is straightforward ("The data on which the information object was released"), but "Termination date" is rather confusing, what does it mean by "termination"? We can have, for example, a codelist that is not valid but we still make it available (for purpose of comparison, etc.)  
    • Attribute "Termination date" might have come from other GSIM object and put under Administrative Details while we were cleaning all attributes during last GSIM revision. It might have made sense with its original object, but now that we put it together with other attributes under Administrative Details, it is not clear. InKyung will check what happened.
    • We should add missing description of attributes while going through objects

  5. InKyung Choi

    14 Dec, 2021

    For discussion on attributes: see GSIM objects with attribute issue