Updated "Home" page for GEDCOM X

[thomast73]

We are making some changes to the GEDCOM X home page and have added a page about the model's relationship to the genealogical research process. I am opening this issue to request feedback on these changes.

[thomast73]

After some further discussion and reading, I have added some additional definition to the genealogical research process page. The graphics likely need to be updated as well.

[thomast73]

Graphics have been updated with the latest feedback.

[jralls]

That first paragraph is awfully dense. Could you open it up a bit, maybe with bullets?
Dr. Jones's doesn't really fit with what you're trying to communicate: I think you really just want the bit bracketed into the "Evidence" bullet.
I didn't like that diagram when you introduced it in #242 and I like it even less here:

• Don't use a class diagram to describe a process.
• Evidence is a subset (not subclass!) of information, the subset that's relevant to the question_. Not sure how to express *that_ in UML; it's more of a database query question.
• The analysis member of Source, Information, and Evidence should be Document, not Analysis.
• There's no mention of the search to find sources.
• Yes, I know that Dr. Jones wrote that "evidence is a tentative answer", and certainly direct evidence is both that and information. The problem is that when making what the lawyers call a circumstantial case, there is no instance of Evidence that provides the tentative answer: Instead, the researcher through analysis infers the tentative answer from a collection of evidence.

There's a stray ')' at the end of the "Questions" sentence.

Every instance of Subject in the (too brief) Modelling section except for Evidence should be Conclusion.

Evidence needs a lot of expansion, starting with how a Subject "contains" information, followed by what to do about Conclusions that aren't Subjects, then how to handle indirect evidence. You also need to address how most genealogists actually work, which isn't n-tier.

[thomast73]

There's a stray ')' at the end of the "Questions" sentence.

Thanks. Fixed. [5383fd5]

That first paragraph is awfully dense. Could you open it up a bit, maybe with bullets?

Agreed. I have attempted a fix. [5383fd5]

Dr. Jones's doesn't really fit with what you're trying to communicate: I think you really just want...

We are trying to describe enough of the genealogical research process that those who are fluent in the process can see the important process concepts and the vocabulary we are using to talk about those concepts. As our description of the research process is very much based in Dr. Jones' descriptions of the process, it is right that we attribute his influence in the material we are publishing here. Referencing Dr. Jones will also be a help to those who are not fluent in the genealogical research process, pointing them to material that can help them learn more.

I didn't like that diagram when you introduced it in #242 and I like it even less here:

• Don't use a class diagram to describe a process.

The UML diagram is not a process diagram. Nor is it a diagram of the GEDCOM X model. It is a data domain diagram. It is meant to represent the essential data concepts the we find as we analysis the data needs of the genealogical research process. The data concepts in the diagram correspond to the concepts described in the research process, and are named such that the concepts in the diagram ought to be easily associated with concepts in the genealogical research process.

• The analysis member of Source, Information, and Evidence should be Document, not Analysis.

Again, the diagram is not a GEDCOM X diagram.

Analysis and Proof are highlighted as distinct data concepts in the research process. It is true that when one models these concepts using GEDCOM X, one would use an instance of Document to represent the data that would be embodied by Analysis and Proof (as suggested), but the diagram is not intended to express this fact.

The section on modeling these data concepts with GEDCOM X data structures is the only section of this document where GEDCOM X classes are intended to appear. This is the section that maps "genealogical research process" data concepts to "GEDCOM X" data concepts.

• There's no mention of the search to find sources.

Again, the diagram is not a process diagram, but a data domain diagram.

We are not (yet) identifying any "search" data entities as "essential" entities that need to be modeled in GEDCOM X.

I tried to add a bit more verbiage to indicate that the UML diagram is a data domain diagram. [5383fd5]

• Evidence is a subset (not subclass!) of information, the subset that's relevant to the question*.
• Yes, I know that Dr. Jones wrote that "evidence is a tentative answer", and certainly direct evidence is both that and information. The problem is that when making what the lawyers call a circumstantial case, there is no instance of Evidence that provides the tentative answer: Instead, the researcher through analysis infers the tentative answer from a collection of evidence.

Information and Evidence are distinct concepts. Evidence is not a subset of Information. Evidence is not a subclass of Information. But even I have been confused into thinking that they are.

Information items suggest answers; they are the building blocks from which Evidence (i.e., answers) are derived. Evidence (answers) only exist in the context of a question. Information exists independent of any question. Given an Information item directly suggesting an answer to a research Question, the instantiation of the Evidence entity (the entity that contains the suggested answer) looks very much like the entity that contains the Information—which might lead one to think that the Evidence item "isa" Information item, or that the Evidence item is just an Information item being referenced from a larger pool of Information items. However, this is not the best way to think about this. As you have stated, it is possible (particularly in circumstantial cases) that none of the Information items resemble the suggested answer—a prime use case for defining Information and Evidence as separate data concepts.

Every instance of Subject in the (too brief) Modelling section except for Evidence should be Conclusion.

No. Subject is the right class here. Research Questions are about identities, relationships or events—i.e., instances of Subject. Informational Conclusions are distinguished from Evidential Conclusions by the extracted property on their containing Subject. Evidential answers refer to their Informational counterparts via the evidence property on Subject.

Evidence needs a lot of expansion, starting with how a Subject "contains" information...

No. If I am modeling Evidence, the Subject contains Evidence, not Information.

You also need to address how most genealogists actually work, which isn't n-tier.

This page is about the genealogical research process as it is defined by those who seek to unify the genealogical community around appropriate research standards (i.e., the Genealogical Proof Standard). However, I do not think we can say that it is a process that "most" genealogists actually use. And while it has a "tiered" sort of feel to it, this document is not a discussion of an "n-tiered" model. The page is strictly focused on the genealogical research process.

[jralls]

Dr. Jones's doesn't really fit with what you're trying to communicate: I think you really just want the bit bracketed into the "Evidence" bullet.

Sorry, that should have been Dr. Jones's slide doesn't really fit...

The new version of the page is much better, especially the UML diagram. But if Analysis is a process, then it should be a method rather than an object.

[thomast73]

...if Analysis is a process, then it should be a method rather than an object.

We are talking about the documentation that results from the processing. I believe that most people would assign the same name ("Analysis") both to the documentation and to processing. So...do want to recommend something different?

[jralls]

Just to be sure we're talking about the same bit, I mean the "Analysis" member of Source, Information, and Answer and the definition of Analysis in the text immediately below that -- which defines Analysis as a process and says nothing about documentation.

Down at the bottom, in the Modelling section you say to use Document "to record analysis and/or proof arguments" (should that be "analyses and proof arguments"? Depends on whether "analysis arguments" makes sense.)

If you don't want to have methods in your diagram, maybe you should change the definition part to explicitly say that it's the written record of the analysis process.

[thomast73]

I have attempted to make the suggested adjustments. I appreciate your insight and feedback. Subtle things, but I feel like they are improvements.

[jralls]

OK, better. "Proof" should be "a conclusion explained in writing" (emphasis only to show the change), as that's part of the 5th GPS element.

Perhaps add a "further reading" section listing the BCG Standards Manual, "Mastering Genealogical Proof", and Christine Rose's Genealogical Proof Standard -- and any others that you can think of, that's just what popped off the top of my head.

On a slightly different subject, should the home page "specifications" sections include pointers to the CV specs (e.g., event-types-specifications) and the Date format spec, or are the hrefs in the Conceptual spec sufficient?

[thomast73]

OK, better. "Proof" should be "a conclusion explained in writing" (emphasis only to show the change), as that's part of the 5th GPS element.

Perhaps add a "further reading" section listing the BCG Standards Manual, "Mastering Genealogical Proof", and Christine Rose's Genealogical Proof Standard -- and any others that you can think of, that's just what popped off the top of my head.

Thanks for the suggestion. I've tried to implement your suggestions (on the wiki side). I look forward to your feedback.

On a slightly different subject, should the home page "specifications" sections include pointers to the CV specs (e.g., event-types-specifications) and the Date format spec, or are the hrefs in the Conceptual spec sufficient?

Our thinking was that, on the front page, we would mention the highest profile specifications and give them a link to see the full specification list on another page. These four specifications seem to be the place most will start, and the other specifications generally augment these four. If we put all of them there, it becomes harder to find the "important" ones (though I hesitate to state it exactly that way). Anyway...that was our thinking.

What are your thoughts in this regard?

[jralls]

Thanks for the suggestion. I've tried to implement your suggestions (on the wiki side). I look forward to your feedback.

Looks good. I think we should be able to come up with some better tutorial material for the further reading -- all of the ones from my initial list aren't really noob-friendly -- but I'll have to look around a bit. It'd be nice to have an online pointer for those too cheap or lazy to get a book.

What are your thoughts in this regard?

Then I suggest a sentence saying that the core concepts are specified in the conceptual spec with a link, followed by another explaining that there are several other pieces and that the whole thing is explained on another page (link). The other page can describe briefly what each subordinate spec contributes, with a link to each.