* Citing Sources: Method 1 and Method 2 - comparison table tweaks

[AdrianBruce]

Adrian,
Can I ask, without getting in trouble, what do the two lump vs split GEDCOM files look like?
I’d want to know how they would translate into my own database.
Thanks

This is the split source-record for Charles W Salter (Source Record #8)

Code: Select all

0 @S8@ SOUR
1 _SRCT @T5@
1 _FIELD EN-REGION
2 ENUM England and Wales
1 _FIELD DT-YEAR
2 DATE 1901
1 _FIELD PL-LOCATION
2 PLAC Bristol, Gloucestershire
1 _FIELD AD-ADDRESS
2 ADDR 7 Wellington Ave.
1 _FIELD NM-HEAD_OF_HOUSEHOLD
2 NAME Charles W /Salter/
1 _FIELD TX-REFERENCE
2 TEXT Class: RG13; Piece: 2400; Folio: 37; Page: 8; Schedule 58
1 _FIELD RP-REPOSITORY
2 REPO @R5@
1 _FIELD TX-COLLECTION
2 TEXT Ancestry "1901 England Census" [database & images on-line]
1 TITL 1901 England and Wales census, Bristol, Gloucestershire, 7 Wellington Ave., Class: RG13; Piece: 2400; Folio: 37; Page: 8; Schedule 58, head of household: Charles W Salter
1 CHAN
2 DATE 24 MAY 2023
3 TIME 21:23:59

And this is the corresponding lumped version for the Ancestry 1901 (Source Record #11)

Code: Select all

0 @S11@ SOUR
1 _SRCT @T6@
1 _FIELD EN-REGION
2 ENUM England and Wales
1 _FIELD DT-YEAR
2 DATE 1901
1 _FIELD RP-REPOSITORY
2 REPO @R5@
1 _FIELD TX-COLLECTION
2 TEXT Ancestry "1901 England Census" [database & images on-line]
1 TITL 1901 England and Wales census, Ancestry "1901 England Census" [database & images on-line]
1 CHAN
2 DATE 24 MAY 2023
3 TIME 21:44:12

The lumped source record is, of course, much smaller. But when you look at the corresponding events, that effect is reversed.

Firstly the census event is cited using split source #8 - the citation stuff for that event is done by just 2 lines (emphasised here by hashes that are NOT in the GEDCOM itself):

Code: Select all

1 CENS
2 DATE 31 MAR 1901
2 PLAC Bristol, Gloucestershire
2 ADDR 7 Wellington Avenue
2 SOUR @S8@		##########
3 QUAY 3		##########

Compare that to the residence attribute cited using lumped source #11. The corresponding citation stuff needs 10 lines, all of which will be repeated in any other "facts" that have the lumped Ancestry 1901 source record, #11, as their source.

Code: Select all

1 RESI
2 DATE MAR 1901
2 PLAC Bristol, Gloucestershire
2 ADDR 7 Wellington Ave.
2 SOUR @S11@		########## starts
3 _FIELD PL-LOCATION
4 PLAC Bristol, Gloucestershire
3 _FIELD AD-ADDRESS
4 ADDR 7 Wellington Ave.
3 _FIELD NM-HEAD_OF_HOUSEHOLD
4 NAME Charles W /Salter/
3 _FIELD TX-REFERENCE
4 TEXT Class: RG13; Piece: 2400; Folio: 37; Page: 8; Schedule 58
3 QUAY 3			########## ends

I want to show this repeating effect in the comparisons but GEDCOM ain't gonna be the way to do it - I'm only adding this here because you asked!

[tatewise]

... Just as it would make sense to resolve whether that Repository should be where the source is (Ancestry) or where the source of the source is (TNA, Kew). I suspect both are needed somewhere.

It is perfectly feasible to have two Repository metafields by modifying the Source Template.
If the Export Gedcom File plugin is used, they will even export to another product as two generic REPOsitory tags which are supported by GEDCOM 5.5.1.

[NickWalker]

Adrian

The two contrasting GEDCOMs are very useful to illustrate the difference (to those who are interested in the way the data is stored), but I think it might help if the second example also used the census fact rather than a residence and the same date. If I remember correctly Adrian you use a Residence fact AND a Census fact for each person recorded in an entry which I don't think is typical of how most people do it. I know this is irrelevant to what you're trying to illustrate but someone might wonder why you've used a census fact of 31 Mar 1901 in one and a residence with date Mar 1901 in the other.

Cheers

Nick

[KFN]

Adrian,

Never mind. Since none of it uses Standard GEDCOM (except TITLE). It does not matter any more, can’t use it in committee discussions!

[tatewise]

KFN, the general picture would be virtually the same for standard GEDCOM.
In the split example, only Source Record tags would hold any details and the Citations would use just the SOUR xref link.
In the lumped example, the Source Record tags would hold fewer details and the Citations would use many of the tags.

[KFN]

KFN, the general picture would be virtually the same for standard GEDCOM.
In the split example, only Source Record tags would hold any details and the Citations would use just the SOUR xref link.
In the lumped example, the Source Record tags would hold fewer details and the Citations would use many of the tags.

I guess I would need an example of this to understand the differences! I just can’t visualize the use of the Standard GEDCOM tags as anything other than lumping!

[AdrianBruce]

... Just as it would make sense to resolve whether that Repository should be where the source is (Ancestry) or where the source of the source is (TNA, Kew). I suspect both are needed somewhere.

It is perfectly feasible to have two Repository metafields by modifying the Source Template. ...

Yes, that would be an alternative resolution to the "which is it?" question.

[AdrianBruce]

... The two contrasting GEDCOMs are very useful to illustrate the difference (to those who are interested in the way the data is stored), but I think it might help if the second example also used the census fact rather than a residence and the same date. ...

That's true, Nick. It just ended up that way since the split source and the citation for the lumped source were deliberately for the same chap to minimise differences in those aspects. But then I had to put them on different "events", otherwise I'd have had the 1901 cited in some form twice against the census event (say).

[AdrianBruce]

... I guess I would need an example of this to understand the differences! I just can’t visualize the use of the Standard GEDCOM tags as anything other than lumping!

Well, here's the equivalent GEDCOM from my "proper" GEDCOM file, which uses non-templated, i.e. generic source records. Firstly the source record (though with most of the Note hacked away and replaced by an ellipsis):

Code: Select all

0 @S4634@ SOUR
1 AUTH General Register Office (England & Wales)
1 TITL schedule for Charles W & Laura M Salter, 7 Wellington Ave., Bristol, 1901 Census
1 ABBR Census Entry: Salter, Charles & Laura, 1901, Bristol
1 _TYPE _Census; Ancestry image
1 PUBL digital image of original published in "1901 England Census [database on-line]"
1 TEXT <<see attached image>>
1 REPO @R24@
2 CALN TNA reference: Class: RG13; Piece: 2400; Folio: 37; Page: 8; Schedule 58
1 OBJE @O3478@
2 _SEQ 1
1 NOTE Search Ancestry English census for Charles White (exact, sounds like, similar) Salter (exact) b 1860 +/- 10y. Use keyword Bristol, rather than birth because sometimes he's just b Bristol, sometimes b Bristol, Somerset (!). 
...
1 CHAN
2 DATE 29 DEC 2021
3 TIME 14:26:05

The Title specifies the exact household and everything leading "up to" the 1901 census. The Published stuff says (effectively) the site and collection. The Call-Number gives the precise TNA reference of that household.

So by the time FH comes to cite that against the Census event, this is the event and all it needs for the citation (again highlighting the citation bit with hashes):

Code: Select all

1 CENS
2 DATE 31 MAR 1901
2 PLAC Bristol, Gloucestershire, England
2 ADDR 7 Wellington Avenue
2 SOUR @S4634@			#### starts
3 DATA
4 DATE 31 MAR 1901
3 QUAY 3			#### ends

Note that there's no "where within source" because I said it all in the Title and the Call-Number.

[NickWalker]

For comparison my Splitter source might look like this (I format the source text differently nowadays - this is an old one). It has a title, source text and a link to an image.

Code: Select all

0 @S429@ SOUR
1 TITL Census 1861 Newton Kyme, Yorkshire 3541/3b (Robert Morley)
1 TEXT Name			Related	Age	Occupation	Birth Place
2 CONT Robert Morley		Head	53	Gardener	Newton Kyme, Yorkshire
2 CONT Rispah Morley		Wife	47	Gardener's Wife	Newton Kyme, Yorkshire
2 CONT Martha Morley		Dau	14	Scholar		Newton Kyme, Yorkshire
2 CONT Mary Anne Morley	Visitor	21			Newton Kyme, Yorkshire
2 CONT Ada Morley Hewitt	Visitor	9 mths			Newton Kyme, Yorkshire
1 OBJE @O313@
1 CHAN
2 DATE 24 NOV 2005
3 TIME 14:00:09

And my citation has nothing at all in it, just the link to the source - in this example the Birth and Census facts for each individual link to the source:

Code: Select all

1 BIRT
2 DATE CAL JUL 1860
2 PLAC Newton Kyme, Yorkshire
2 SOUR @S429@                              #### <---- 1st Citation
1 CENS
2 DATE 7 APR 1861
2 PLAC Newton Kyme, Yorkshire
2 SOUR @S429@                             #### <---- 2nd Citation

[NickWalker]

For comparison my Splitter source might look like this (I format the source text differently nowadays - this is an old one). It has a title, source text and a link to an image.

Code: Select all

0 @S429@ SOUR
1 TITL Census 1861 Newton Kyme, Yorkshire 3541/3b (Robert Morley)
1 TEXT Name			Related	Age	Occupation	Birth Place
2 CONT Robert Morley		Head	53	Gardener	Newton Kyme, Yorkshire
2 CONT Rispah Morley		Wife	47	Gardener's Wife	Newton Kyme, Yorkshire
2 CONT Martha Morley		Dau	14	Scholar		Newton Kyme, Yorkshire
2 CONT Mary Anne Morley	Visitor	21			Newton Kyme, Yorkshire
2 CONT Ada Morley Hewitt	Visitor	9 mths			Newton Kyme, Yorkshire
1 OBJE @O313@
1 CHAN
2 DATE 24 NOV 2005
3 TIME 14:00:09

And my citation has nothing at all in it, just the link to the source - in this example the Birth and Census facts for each individual link to the source:

Code: Select all

1 BIRT
2 DATE CAL JUL 1860
2 PLAC Newton Kyme, Yorkshire
2 SOUR @S429@                 #### <---- 1st Citation
1 CENS
2 DATE 7 APR 1861
2 PLAC Newton Kyme, Yorkshire
2 SOUR @S429@                 #### <---- 2nd Citation

[KFN]

Thanks all,

Pushing what I think of as “source detail” up to the REPO.CALN has some value if the REPO actually represents the census rather than the holder of the census.

Example of what I normally do with a census (Lumping):
1) Repository_Record = Holder of the census documents (library, A.com, etc)
2) Source_Record = Census (1890 Federal US, 1895 Bergen, Norway), REPO.CALN details where_in_Repo the census is stored.
3) Source/Citation Detail = Where_In_Source (SOUR.PAGE)

From Adrian:

Code: Select all

0 @S4634@ SOUR
1 AUTH General Register Office (England & Wales)
1 TITL schedule for Charles W & Laura M Salter, 7 Wellington Ave., Bristol, 1901 Census
1 ABBR Census Entry: Salter, Charles & Laura, 1901, Bristol
1 _TYPE _Census; Ancestry image
1 PUBL digital image of original published in "1901 England Census [database on-line]"
1 TEXT <<see attached image>>
1 REPO @R24@
2 CALN TNA reference: Class: RG13; Piece: 2400; Folio: 37; Page: 8; Schedule 58
1 OBJE @O3478@
2 _SEQ 1
1 NOTE Search Ancestry English census for Charles White (exact, sounds like, similar) Salter (exact) b 1860 +/- 10y. Use keyword Bristol, rather than birth because sometimes he's just b Bristol, sometimes b Bristol, Somerset (!). 
...
1 CHAN
2 DATE 29 DEC 2021
3 TIME 14:26:05

What I’m feeling (Splitting):
1) Repository_Record = Census (1890 Federal US, 1895 Bergen, Norway)
2) Source_Record = Specific page of the census, REPO.CALN details where_in_Repo (the census) the detail is found
3) Source/Citation Detail = blank

[KFN]

For comparison my Splitter source might look like this (I format the source text differently nowadays - this is an old one). It has a title, source text and a link to an image.

Code: Select all

0 @S429@ SOUR
1 TITL Census 1861 Newton Kyme, Yorkshire 3541/3b (Robert Morley)
1 TEXT Name			Related	Age	Occupation	Birth Place
2 CONT Robert Morley		Head	53	Gardener	Newton Kyme, Yorkshire
2 CONT Rispah Morley		Wife	47	Gardener's Wife	Newton Kyme, Yorkshire
2 CONT Martha Morley		Dau	14	Scholar		Newton Kyme, Yorkshire
2 CONT Mary Anne Morley	Visitor	21			Newton Kyme, Yorkshire
2 CONT Ada Morley Hewitt	Visitor	9 mths			Newton Kyme, Yorkshire
1 OBJE @O313@
1 CHAN
2 DATE 24 NOV 2005
3 TIME 14:00:09

And my citation has nothing at all in it, just the link to the source - in this example the Birth and Census facts for each individual link to the source:

Code: Select all

1 BIRT
2 DATE CAL JUL 1860
2 PLAC Newton Kyme, Yorkshire
2 SOUR @S429@                              #### <---- 1st Citation
1 CENS
2 DATE 7 APR 1861
2 PLAC Newton Kyme, Yorkshire
2 SOUR @S429@                             #### <---- 2nd Citation

This make some sense as well, although it does not use the AUTH, PUBL tags for citation completeness. It holds closer to my understanding of GEDCOM.

Obviously, future GEDCOM needs more normalization, separating out both "Facts" and Citation Detail. I not sure if I propose that to "The Committee" I'll get much traction, but that is what I would do. "Normalize, Normalize, Normalize:!!!

[NickWalker]

This make some sense as well, although it does not use the AUTH, PUBL tags for citation completeness. It holds closer to my understanding of GEDCOM.

Obviously, future GEDCOM needs more normalization, separating out both "Facts" and Citation Detail. I not sure if I propose that to "The Committee" I'll get much traction, but that is what I would do. "Normalize, Normalize, Normalize:!!!

No I have no need for those fields so I don't use them but I could if I needed to of course. Regarding normalisation, as a database designer, this is precisely why I use split sources rather than a lumped source because I don't want the duplication that using citations brings. I know I'm repeating what others have said but I want to transcribe my census entries and this would have to be recorded in the citations if I used a lumped source. But as a typical census household would generate 10 or more citations, I'd end up having to duplicate the same transcription 10 times (using a shared note linked to the citation instead would be possible but it does complicate things). Even if I wasn't transcribing and I recorded things like the 'Where in Source' in the citation I'd again have to duplicate that into the 10+ citations and if I was to get that reference wrong I've made the same mistake 10 times and have to change it in all those places. I really do dislike any method that requires the same data to be recorded multiple times. The splitter source method gets away from that and makes it closer to a normalised database model that I like. It works very well and I suspect it has become particularly popular for a section of FH users because it is the method that my Ancestral Sources (FH companion product) has used as the default for nearly 20 years.

If a future GEDCOM does ever build in a shared citation structure to avoid all this duplication then that would be great!

[tatewise]

If a future GEDCOM does ever build in a shared citation structure to avoid all this duplication then that would be great!

The alternative would be for FH to support a 'shared citation' feature in the user interface, as do many other products.
Then editing the 'replicated citations' would not be such an issue. Yes, the citation-specific data would still be replicated in the GEDCOM file (unless FH introduced a custom structure) but it would not impact the user as much.

[AdrianBruce]

... if I wasn't transcribing and I recorded things like the 'Where in Source' in the citation I'd again have to duplicate that into the 10+ citations and if I was to get that reference wrong I've made the same mistake 10 times and have to change it in all those places. ...

To illustrate what Nick says, some more screenshots etc comparing the 2 methods for (nearly) the same people. Hopefully a picture is worth a thousand words.

I have taken four baptisms at Barthomley - Mary in 1777 and Betty in 1779 to Samuel Billington; with Martha in 1781 and Rebecca in 1783 to Samuel Cooper. The text in the original register is always a variation on

Baptisms 1783
Rebecca Daughter of Samuel Cooper of Barthomley Taylor March 19th

The baptisms for Martha and Rebecca Cooper were entered first using Split / Method 1 source-records, i.e. one source-record per event. Full details were in their corresponding source-record.

The baptisms for Mary and Betty Billington were then entered using a single Lumped / Method 2 source-record that represented all parish registers for Barthomley. The specific details for their two baptisms appear in the citation details, as expected with this method, rather than the source-record(s).

The image below shows the citation window for the baptism of the first split source child.

Screenshot 2023-05-27 113024.jpg (106.09 KiB) Viewed 2588 times

The image below shows the citation window for the baptism of the first lumped source child.

Screenshot 2023-05-27 113138.jpg (100.11 KiB) Viewed 2588 times

Martha's (split) source-record generates 9 citations, all containing only the Assessment of "Primary Information".

Rebecca's (split) source-record generates 6 citations, all containing only the Assessment of "Primary Information".

Martha's extra 3 citations are (1) the whole record citation for her father, (2) the citation for his name and (3) the whole record citation for her parents' family.

This makes 15 citations using the split-source method.

The lumped source-record for Barthomley parish registers has 15 citations, 9 associated with the input of Mary's baptism and 6 associated with Bettey's. This shows that in this case certainly, the total number of citations is unchanged whichever method (Splitting or Lumping) is chosen.

What are some other consequences of the choice?

1. Firstly note the entirely accidental mis-spelling of St. Bertoline as "St. Beroline" on the lumped source record. Because this mistake is on the source record, it can be fixed with one edit to the source record.

2. The first time I entered this data, I (again accidentally) forgot to add the Text From Source to the citation associated with Mary Billington's baptism. There are nine such citations in the final file, all of which have to be corrected individually. Yes, the citations are identical but FH has no means of linking them.

This illustrates one major issue with lumping in FH (referred to above by Nick) - any modification to citation level data needs to be done multiple times - lumped sources have lots of citation level data. Conversely, split sources have virtually all their data in the source record, so that particular fix of the missing Text from Source (say) would only be done once, on the source record.

Note that correcting the error with the church name on the lumped source is a single correction but if the data had been lumped differently, e.g. all parish registers on one source record, then that church name fix would also need to be done multiple times down at citation level.

3. Finding the citations associated with Mary Billington's baptism to correct.
(a) Go to the Source Record concerned (the one for the Barthomley PRs)
(b) Use option Cog-Wheel / Show Source Record's Citation in Result Window.
See screen shot (which may need magnifying, e.g. by opening the image in a new tab) for the result

Screenshot 2023-05-27 115641.jpg (293.82 KiB) Viewed 2588 times

(c) Locate the Mary Billington citations in the column Citation (Footnote Text). Double click the entry in the Citation column on that line item, find the right source in the resulting window, and edit the Source Citation. (Other routes are available from that window).

[ColeValleyGirl]

I'm not ignoring this discussion, but need to consider how best to use the material being generated.

[AdrianBruce]

I'm not ignoring this discussion, but need to consider how best to use the material being generated.

Absolutely agree...