clarify how proof of a person is connected to proof of a fact #120
Comments
|
Thanks for bringing up the question. This needs to be clarified. I've been meaning to get around to that for awhile now. You're not the first to bring it up. I'll try to tackle this for 0.10.0. |
|
My take on this is that a Person (Conclusion Model) is primarily a collection of source references where the Persona has been researched to deduce whether or not it is thought to match the Person being researched. The collection should include negative results as well as positive ones (ie should include why Persona1 is not thought to be Person ABC as well as why Persona2 is thought to match Person ABC). From each Persona we/the application can track the Facts and Relationships from the Records so it is not essential to re-create these in the Conclusion Model. The researcher typically (and selectively) builds the life event(s) of a Person from the Facts/Relationships/Persona but these do not necessarily correlate directly. For example:
Conclusion Model might look like this: Person CP1: Person CP2: Supposing we discover that R3.P1 is in fact a different person ... so R3.P1 changes from being a positive assessment to a negative assessment in the collection against CP2. Providing that this is done at the Person level (and not just the Fact level) we can then deduce and highlight what needs to be re-assessed (ie all Facts for CP2 which also reference R3.P1). We can also deduce that since R3.P1 is in a Relationship with R3.P2 then all references to R3.P2 are now also suspect. Hence we can highlight that CP1's death is suspect. All this, depends on: (a) holding a separate evaluation of the Person/Persona match for every resource referenced by a Fact etc for a Person. If we break this and allow a fact to reference a source without an evaluation of the Person match then the reference is ambiguous - does the Proof Statement refer to the Person/Persona match or to an assessment of the truth of the Record.Fact? However, source references which do not link to any Persona do not need to be held/evaluated at the Person level (since they don't have a relevant Persona to link to). In the example above I used an illustration of a Place. If the illustration was later found to relate to another Place then this would be changed at the level of the Place entity and anything which referenced R4 could be highlighted as dubious. |
|
An alternative way of describing Persons could be via a Wiki-type narrative rather than using Facts at all. In the above example, we could have: Person CP1: Person CP2: For Persons whose narrative has not been created (say because the Person is a distant leaf in the researcher's tree), the Record model can be used to show what has been linked without the need for any Facts to be created by the user, e.g.: Person CP1: In any of the above cases the Person.Facts are optional and are at the discretion of the software being used, but the Person/Persona links in the data model are critical. |
|
I'm starting to take this issue on with consideration that the record model isn't "visible" from the conclusion model anymore, except as just another derivative source. So I was thinking that a simple clarification would be to state that the sources of all conclusions (gender, name, fact) must refer to sources declared at the level of the of the person. Then I'd provide some recipes to illustrate this. Would this be an adequate clarification? What's missing? |
|
Not sure ... I think what I'm needing is a direct Person->Persona link so that there is a connection not just to the source but to the Persona referenced within it. However, the 'Persona' I am referring to is one constructed by (or at least editable by) the researcher not necessarily by the vendor of the Record Model (something I was trying to understand/explore in #151) For me, there is a distinction between a source cited as Evidence (ie a source which is intended to contribute towards proof of the Person/Fact) and a source which is cited as an illustration (e.g. a citation from an article on what life was like in a workhouse or a description of a place etc) - see my note in #141. Hence, in my model I have an Evidence object which links to the Persona; and a (separate) Citation which is just a link to the source. The Evidence can be validated as I illustrated above; the citation can't be validated since we don't even know if the Person was mentioned - it's effectively just a footnote to credit the author(s). I suspect this may be too much detail and too specific an implementation for others so I'm happy to compromise but could we explore the concept of a Person->Persona link before we throw it out? |
|
I think I'm confused. Can you clarify exactly what you would need in order for this issue to be considered "closed" in your mind? Would some documentation that explains how to distinguish between a source cited as Evidence and a source cited as an illustration be adequate? |
|
Oh dear :( I'm sorry. It may well be that this should be an application-specific thing rather than in the standard so maybe the short answer is for you just to say "No, we won't do this." But to give it one more go ... My requirement would be for concluded Persons to have a collection of Evidence objects, each Evidence object comprising the following properties:
Does that make sense? |
Not at all. I appreciate your patience with me.
Yes. Thank you. I can do that. |
|
Phew! Many thanks :) |
|
Hey, @EssyGreen, please consider the recipes that have been added to the recipe book to provide for the cases you've asked about. What other questions, issues, or concerns still remain on this topic? |
|
I've been awol for a while - just getting back into it. Give me a while to catch up on the new stuff - I'll comment asap. |
|
I still can't find any examples of negative proof or any examples which allow the researcher to explain why the citation is thought to refer to (or not refer to) the relevant person. All the examples seem to be falling into the common trap of "I found it therefore it must be true". I understand the need for simple examples but think there is a desperate need to exemplify the importance and complexity of deeper investigation and analysis. |
+1! To expand a bit, inference is critical to good genealogy, and GedcomX needs to be able to document inferred conclusions which depend on indirect evidence from several sources -- including the absence of evidence. |
Indeed. Not modeled yet. That's tracked at #127.
Indeed. I'll add that concept to the examples. My bad that it got left out.
That's fine. Thanks for your patience. I suffer from being too close to the project to understand what it's like for people who are approaching it from a distance. Things that are clear in my head are hard for me to articulate in a way that it can be clear for everybody else. I'd be interested to know how you'd like to see things articulated. Would more of a narrative help? What if the recipes included some text that told a story? E.g. "Sarah is doing research on George Washington and found this source that included information about his birth date and she does some analysis, etc., etc. etc." Would you be willing to write some narratives for me and I can fill in the technical details? |
|
@stoicflame - I suffer from being awol for too long hehe ... It's actually interesting coming back to it afresh ... which may give some insight into articulation ... as an "outsider" coming back it it feels that there is a very detailed technical format there but very little in terms of overview and/or summary to lead you in ... you go from a high level diagram slap bang into XML/RDF/FOAF definitions and it's really difficult to see (a) how it hangs together and (b) how to use it. This applies to both the conceptual spec and the recipe book. Also, (and perhaps more importantly) it feels that its all about the technology and we've lost the genealogy. Taking the baking a cake example ... Here I am in my genealogy kitchen ready to make my cake ... I'm already a cook so I understand the basics of cake making but I've just bought my GEDCOM X Cookbook 'cos it was supposed to make the most wonderful cakes and get around all those pesking problems of the cake sinking in the middle or getting burnt. So I open it up and start reading ... There are some nice pictures of a couple of generic cakes just inside the cover (=model diagrams) so I'm encouraged to read on ... But the bulk of the book is loads of technical information about the equipment I need to use (=detailed definitions in the spec) .. It goes on about exactly the size of the baking tins, the thickness of the grease-proof paper, what make of oven I must use and the health and safety equipment I must have in my kitchen. ... Er I think I'll skip that part. Right here's a recipe (=recipe in the recipe book) ... er ... it doesn't mention the ingredients ... it just gives the chemical compound of the finished cake!! Eesh I guess I can make out that there was some flour, sugar and currants in there somewhere but where did the eggs go? What would I like to see? More narrative? Not necessarily - more genealogy - real genealogical data/situations are needed. A "How Do I..." approach e.g. "... enter a simple BMD certificate", "Record a Census record which includes unrelated people in the household", "Enter the possible matches for an IGI record", "Record the secondary roles mentioned on a marriage certificate", "Transcribe a detailed Will where some of the bequests are illegible", "Log my search results across multiple repositories", etc What is missing for me is the "method" part of the recipe - or to put it another way the process part of genealogy ... e.g. You currently have: "The following example illustrates how to cite an online record. Evidence for Israel Hoyt Heaton is found in the 1920 U.S. Census. The URI to the record is "https://familysearch.org/pal:/MM9.1.1/M8PT-4GN". The URI for a description of the record is "https://familysearch.org/platform/sources/GGG-GGGG"." You seem to start with an existing person: And a source record: 1920 US Census https://familysearch.org/pal:/MM9.1.1/M8PT-4GN (no idea what is in the source record tho and presumably no where to record this in GEDCOM X!!!!) These seems to end up as: What did this mean? What did the researcher do? It looks like they just happened to find a guy with the same name in the 1920 Census and went "Yes it must be a match I'll add it as a citation" and they bunged the reference at the end of the record to "prove" it. How about this instead:
How does this fit with GEDCOM-X? I don't know ... and that's my problem ... it's not the syntax which is a problem - it's the model. Would I be willing to write some? I'd love to but time as always is the problem ... How about starting a thread for use cases that we can all add to and you can pick from these to code up and include in the recipe book? |
|
Excellent comments. Like you said, these kinds of things take a lot of time to put together, so they have to be prioritized according to availability of resources. It should be no surprise that the team working on GEDCOM X is very much focused on development and technical details. And we're also not working on just GEDCOM X, but also the new FamilySearch Developer Platform. So while we're very strong on the technical details and real, working code, we fall short in our abilities to gather all of the "real genealogical data/situations" that you express a need for. We've got a set of requirements for FamilySearch application(s), but that's just one vendor's perspective on the world.
Can we not just use the issue tracker? Open up a new issue for each of those different scenarios? |
|
Okay. I'm thinking it would be nice to create a "template" so others can know what kind of details we need to create a real recipe. I'm thinking there might be other slow people like me that don't know how to write a recipe and I guide or a template could help. I wonder, could we leverage a Google Doc (document or spreadsheet)? Or maybe the BetterGEDCOM wiki? Would either of those solutions work better? I'm trying to think of the most efficient way to organize the effort. |
|
Uh, there's already a wiki provided by Github. Why not use that? I don't think it's fair to the BetterGedcom folks to barge in on theirs, though it would be nice to invite them to contribute to yours. Yank the recipes.html page and redo it as a top-level page in the GedcomX wiki, pasting in the list from recipes.html as a starting point. Post an announcement on your blog asking folks to try to figure out from the spec how to do something and to create a recipe documenting it... and to write an issue if they can't figure it out. |
+1 |
|
Okay, let's try it out. Would you guys be willing to create a wiki page that looks like an actual recipe so we can put together some clear instructions that others can follow? |
|
Absolutely... once I fully grok the spec! In particular, I need an answer to the questions about precedence, etc. I asked yesterday in #165 and which got buried in the discussion about RDF. |
Oh, I didn't want to impose that as a prerequisite :-). I was just hoping you could come up with the text, inputs, outputs, expectations, scenarios, whatever needs to go in. Leave big holes for where the code needs to go and I'll fill those in (if possible). |
|
@stoicflame - Give me a place to do it and I'll detail some cakes I need recipes for :) |
|
OK, I can do that. |
The proof of a Fact which is related to a Person or Relationship is totally dependant upon the proof that the Persona(s) are proven to be the same Person(s). If any evidence (maybe discovered at a future date) disproves that Persona ABC relates to Person XYZ then all concluded facts which are based on this assumption are null and void. How can this be ascertained given that there is no connection between the Attribution in the Fact and that of the Person?
The text was updated successfully, but these errors were encountered: