Commons talk:Structured data/Modeling/Date
These earlier notes and resources may be inspiring:
- Notes from first data modelling discussion at Wikimania 2019 - date modelling was discussed here, and the discussion is recorded in the notes.
- Notes from second data modelling discussion at Wikimania 2019
- Properties table - already contains some first ideas on date modelling
- Interesting Wikimedia Commons files collected because of their structured data modelling challenges
– SandraF (WMF) (talk) 10:34, 4 September 2019 (UTC)
'Time' data type not active yet
[edit]Please note that at the time of writing this comment (September 4, 2019), the 'Time' data type is not active yet on Wikimedia Commons. It is planned to be activated in the upcoming months though. SandraF (WMF) (talk) 11:29, 4 September 2019 (UTC)
Date notes from Wikimania 2019
[edit]Copy here for future reference
- Wikidata (publication date (P577) = publication date; production date (P2754) = production date ; inception (P571) = inception)
- Date of creation (inception (P571))
- Creation of what? The file? What type of creation? When a photo was taken?
- Date of publication (publication date (P577))
- Date of upload (publication?)
- Date of modification
- Date of destruction (of digital copy if file is destroyed)
- Start and end date if applicable
- Date properties should be direct to be able to add qualifiers
- Qualifiers “earliest date (earliest date (P1319)), latest date (latest date (P1326)), refine date (refine date (P4241)),
- How to tell user (like another site) which date they should show?
End of copy Multichill (talk) 18:25, 12 September 2019 (UTC)
Here the question is, what is behind the concept of "depicts of depicts". That may answer whether to use the same property for file and depicted object creation time or not. Juandev (talk) 08:27, 17 September 2019 (UTC)
- Humm... we use already at Wikidata the "significant events", I do believe that this is the correct path follow.
- If there is a necessity, we can include all "modifications", "date of destruction", "start and end date", ... as we do it now at Wikidata.
- And is kind hard to brainstorm all the possible dates.
- I'll give an example: Notre-Dame de Paris, is quite valuable to have all important events that happened to this cultural heritage collapsed at the file description, is not that important see all files modifications,tough.
- And we already have that at the historic of the previous versions.
- I would prefer import all the dates collapsed at the description simply importing "significant events".
- -- Rodrigo Tetsuo Argenton m 22:02, 2 October 2019 (UTC)
- The mayor issue is that there are only days active on Wikidate. But we also need hours minutes and seconds. And then we need to manage time zones.--GPSLeo (talk) 07:30:02, 3 October 2019 (UTC)
-
- GPSLeo,
- The precision at Wikidata goes of second to billion years
- Property talk:P2803
- Also, we can add a qualifier to include the time zone.
- -- Rodrigo Tetsuo Argenton m 08:10, 3 October 2019 (UTC)
- The software supports this but it is currently disabled. Time zone as qualifier could work but is hard to query. --GPSLeo (talk) 08:39, 3 October 2019 (UTC)
-
- So may request to rearm this, seems to be the easiest way.
- And not that difficult to query... -- Rodrigo Tetsuo Argenton m 09:10, 3 October 2019 (UTC)
Start with simple dates
[edit]I noticed Rudolphous adding categories like Category:Netherlands photographs taken on 2010-09-25 like for example on File:Haarlem - Bakenessergracht 100.JPG. The dates on these images have very straightforward formatting (YYYY-MM-DD) and we can use it to add inception (P571). I don't expect more precise date time any time soon. Multichill (talk) 13:17, 9 November 2019 (UTC)
- To input time, there is refine date (P4241) and a series of items (d:Wikidata:WikiProject Calendar Dates/lists/time of the day). Obviously, nothing one would want to add manually. Jura1 (talk) 09:08, 20 November 2019 (UTC)
- I think we should wait till there is a clear way to store all this metadata. If we do this without a clear plan we will have 100 millions of botedits. As we have much more images here then there are articles on Wikipedia this is unlikely more huge then the growth of Wikidata. --GPSLeo (talk) 10:45, 20 November 2019 (UTC)
- I think I will start adding inception (P571) to my files. I would only add dates with day precision as I do not see higher precision being available in near future. I really dislike refine date (P4241) being used to store time of day as that is just such a hack. --Jarekt (talk) 04:07, 20 December 2019 (UTC)
- I think we should wait till there is a clear way to store all this metadata. If we do this without a clear plan we will have 100 millions of botedits. As we have much more images here then there are articles on Wikipedia this is unlikely more huge then the growth of Wikidata. --GPSLeo (talk) 10:45, 20 November 2019 (UTC)
inception
[edit]Why is inception the most relevant? Inception is that date something began to exist. For a photograph, than might be the date the photograph was taken or the date the photograph was printed, or the date it was scanned, or the date is was digitized. It might be the date that the object in the photo came into existence. For a scan of a book, that might be the date the author started writing the book, or the date of first publication, or the date the scan of the book was created.
An unqualified and unclarified "inception" is of no value. --EncycloPetey (talk) 14:11, 2 April 2020 (UTC)
- We are following norms used on Wikidata, and for photograph inception is the date photograph was taken. The dates or photograph printing or scanning are not relevant. In case of a book we should use publication date (P577). --Jarekt (talk) 17:53, 3 April 2020 (UTC)
- In case of book the wikitext should use for example {{Date context|published|1908}} --Schlurcher (talk) 08:01, 4 April 2020 (UTC)
- Yes, we use inception (P571) for when a photograph was taken. If you want to record when it was first published, you can use publication date (P577). For other significant events you can use significant event (P793) with qualifier point in time (P585). See also d:Wikidata:WikiProject_Visual_arts/Item_structure#Describing_individual_objects. Multichill (talk) 13:36, 4 April 2020 (UTC)
- To me it sounds much more naturally that a photo is a snapshot with a specific point in time (P585), rather than a date of inception. --Jonatan Svensson Glad (talk) 19:21, 12 April 2020 (UTC)
- We have established ways of modeling different metadata, which were discussed and argued over for for a decade on Wikidata. We are not going to establish different ways of modeling basic properties, because "it sounds much more naturally". --Jarekt (talk) 13:37, 31 August 2020 (UTC)
- Well, this is not Wikidata. This is Commons and we have not discussed this at all. --Jonatan Svensson Glad (talk) 19:55, 14 September 2020 (UTC)
- We did, you just didn't bother to participate. We're no idiots, we didn't start from scratch. We're improving on the model that was started on Wikidata. Multichill (talk) 20:41, 14 September 2020 (UTC)
- Please also note that the item we are using is only P571, a number. The number renders differently in different languages. The German rendering is more like "date of creation", which fits quite nicely to the time a photo was created. So if at all, the English label could be improved, not necessarily do we need to change the number identifier. What would be your preferred English transcription? --Schlurcher (talk) 07:13, 15 September 2020 (UTC)
- We did, you just didn't bother to participate. We're no idiots, we didn't start from scratch. We're improving on the model that was started on Wikidata. Multichill (talk) 20:41, 14 September 2020 (UTC)
- Well, this is not Wikidata. This is Commons and we have not discussed this at all. --Jonatan Svensson Glad (talk) 19:55, 14 September 2020 (UTC)
- We have established ways of modeling different metadata, which were discussed and argued over for for a decade on Wikidata. We are not going to establish different ways of modeling basic properties, because "it sounds much more naturally". --Jarekt (talk) 13:37, 31 August 2020 (UTC)
- To me it sounds much more naturally that a photo is a snapshot with a specific point in time (P585), rather than a date of inception. --Jonatan Svensson Glad (talk) 19:21, 12 April 2020 (UTC)
- Yes, we use inception (P571) for when a photograph was taken. If you want to record when it was first published, you can use publication date (P577). For other significant events you can use significant event (P793) with qualifier point in time (P585). See also d:Wikidata:WikiProject_Visual_arts/Item_structure#Describing_individual_objects. Multichill (talk) 13:36, 4 April 2020 (UTC)
- In case of book the wikitext should use for example {{Date context|published|1908}} --Schlurcher (talk) 08:01, 4 April 2020 (UTC)
Well this matter seems firmly settled but I just thought I'd point out, for other editors who end up here, that there's also the date depicted (P2913) property which could apply to photographs. --▸∮truthious ᛔandersnatch◂ 14:13, 17 October 2020 (UTC)
- date depicted (P2913) sound more fitted to the situation where within the photograph there is a date depicted. Like if you take a picture of a calendar. --Schlurcher (talk) 15:11, 17 October 2020 (UTC)
- Agree with Schlurcher. Date depicted would be for a sign with a date, or a ticket stub, or a hat worn at New Years that shows the year. Senator2029 22:06, 17 October 2021 (UTC)
Often, I see where upload date is recorded in the inception field, when the image (a logo, map, etc.) was first created at some time prior. The rationale to use word "inception" is faulty and needs to be reconsidered. To me, the word doesn't mean "date the photo was taken", it means "date the subject of the photo came into existence". This is how is inception is used on Wikidata: to record when an organization was established. To record the date a work is created, properties like date published are used, not inception. So why do we have is backwards here on Commons. And point in time is best for photographs, as a photograph is an image only at one point in time. For photographs, stop using "inception" and use "point in time" instead. Senator2029 21:58, 17 October 2021 (UTC)
Exif date
[edit]I noticed quite a few files still in Category:Images from Wiki Loves Monuments missing SDC inception. Some of these files are like File:052 2015 06 16 Kulturdenkmaeler Forst.jpg where {{Taken on}} is used and I have to adapt the bot, but quite a few others don't have a suitable date in the template. On File:1 Kotsiubynskoho Street, Lviv (04).jpg no date and on File:1022 Kalhokstr 1-3-5 37693.JPG (which happens to be the date of creation). I'm considering adding the date from exif in these cases. Multichill (talk) 10:27, 30 August 2020 (UTC)
- Just to put this a bit in perspective: Currently we have 2.4M WLM uploads of which about 52K don't have inception (P571) so that's only about 2.5% to do or 97.5% done. But WLM uploads are relatively clean because most are done using stand upload tools, more benefit in other uploads. Multichill (talk) 11:02, 30 August 2020 (UTC)
- I would first add {{According to Exif data}} to the file description. We should then model them the same way as we would do with all files with this template. I do think, however, we do not have a plan for this template, yet. --Schlurcher (talk) 18:07, 30 August 2020 (UTC)
- Maybe the bots of @Aschroet: Commons:Bots/Requests/ArndBot or @TheSandDoctor: Commons:Bots/Requests/TheSandBot can help with the first task. They seem to have bots approved for this or a similar task. --Schlurcher (talk) 08:08, 31 August 2020 (UTC)
- I would first add {{According to Exif data}} to the file description. We should then model them the same way as we would do with all files with this template. I do think, however, we do not have a plan for this template, yet. --Schlurcher (talk) 18:07, 30 August 2020 (UTC)
There seem to be a great many reasons for lack of date and each might need a different solution:
- No date in {{Information}} but there is one in EXIF. Example: File:2 Shukhevycha Street, Lviv (03).jpg. Solution add {{According to Exif data}} and reprocess
- file uses {{Photograph}} or {{Artwork}} instead of {{Information}} infobox; Example File:Birkenfeld, Kirchgasse 2-20160506-003.jpg or File:Francesco francia, adorazione del bambino con santi alla presenza di anton galeazzo e alessandro bentivoglio, 1498-99, da s.m. della misericordia 05.jpg. Solution: process the dates and accept if they are >2000
- date not in YYYY-MM-DD format: but uses YYYY or YYYY-MM format. Example File:Bliesheim Frankenstraße 61 01.JPG or File:Gossam 7730.jpg. Solution: process dates in that format
- date in correct YYYY-MM-DD format. Example File:Huis Lievendaal 9.jpg or File:Gossam 7730.jpg. Solution: reprocess
- Date preprocessed through some template:
{{Date|2010|09|26}} (upload date)
; Example: File:Amersfoort Kamp 48.JPG. Solution add {{According to Exif data}} and reprocess Done{{According to EXIF data|2016-08-20}}
; Example: File:Gortz kirche baudenkmal - 7.jpg. Solution: add code to recognize and reprocess{{Taken on|07.08.2018 14:15:39}}
; Example: File:Bad Bergzabern Obere Berggasse 06 004 2018 08 07.jpg. Solution: add code to recognize and reprocess{{ISOdate|2014-09-01 19:03:53}}
; Example: File:Die Ginsburg.jpg. Solution: add code to recognize (or just remove {{ISOdate}}) and reprocess{{between|2011|2012}}
; Example: File:(004) 9-008 Jüdischer Friedhof, Am Dormannsbusch (Neuss-Grimlinghausen).jpg. Solution: get EXIF or encode as is{{original upload date|2010-09-18}}
; Example: File:Gageldonk.JPG. Solution add {{According to Exif data}} and reprocess Done2019-10-09 {{wrong date}}
. . Solution add {{According to Exif data}} and reprocess
- users use personalized infobox templates, which then call {{Information}} template. Example File:Foutain Girl with a Jug in Tsarskoe Selo 01.jpg or File:Ehemaliger Gasthof Hirschen (Liptingen) Detail 2.jpg. Solution: subst the template and replace with standard infobox template or write bot that analyses html.
Majority of files do not have a date because we are skipping files with {{Photograph}} or {{Artwork}}. If we extract dates from those first there will be much less fines to look through. --Jarekt (talk) 13:32, 31 August 2020 (UTC)
- I like the two step approach of extracting the data and putting it in standard templates and another bot picking it up again for structured data. Not a lot of people are comfortable with extracting dates and working with structured data (I at least are a bit reluctant about the date part). @Rudolphous: do you think you can help out with the date extraction and clean up? Multichill (talk) 16:15, 31 August 2020 (UTC)
- Doing the "original upload date" one right now, for example [1]. There are around 400 files and a very few are skipped because there are no dates in exif. Rudolphous (talk) 19:59, 31 August 2020 (UTC)
- Ok, great. This search (98.047 hits) will give some possible suggestions. I assume you found the exif part of the api accessible in Pywikibot as filepage.latest_file_info.metadata . Multichill (talk) 20:16, 31 August 2020 (UTC)
- @Schlurcher: I could definitely help out with exif dates if still needed. Just let me know what needs doing & an example. --TheSandDoctor (talk) 05:05, 1 September 2020 (UTC)
- @Multichill: Also the "Date" (first one) is done (around 3000). Handy api, didn't know this one. I don't use python btw. Rudolphous (talk) 06:00, 1 September 2020 (UTC)
- @Multichill: Also thanks for the search. Added a resultcounter to see in time if this goes up or down. Rudolphous (talk) 06:09, 1 September 2020 (UTC)
- Thanks. Good to see the team work. --Schlurcher (talk) 07:08, 1 September 2020 (UTC)
- @Rudolphous: still doing date updates? My example links was just for Wiki Loves Monuments. Should be plenty more files needing attention. Files like File:Gebouw van Eindhoven Airport.jpg for example should be updated too. Probably best to focus on easy bulk cases. At some point in the (more distant) future we can add a tracker category to {{Information}} so it's easier to find files that don't have a date. Multichill (talk) 19:22, 7 September 2020 (UTC)
- Thanks. Good to see the team work. --Schlurcher (talk) 07:08, 1 September 2020 (UTC)
- Ok, great. This search (98.047 hits) will give some possible suggestions. I assume you found the exif part of the api accessible in Pywikibot as filepage.latest_file_info.metadata . Multichill (talk) 20:16, 31 August 2020 (UTC)
- Doing the "original upload date" one right now, for example [1]. There are around 400 files and a very few are skipped because there are no dates in exif. Rudolphous (talk) 19:59, 31 August 2020 (UTC)
- @Multichill and Jarekt: , how do we envision to model {{According to Exif data|2010-09-05}}. Just the date or add some qualifier? --Schlurcher (talk) 07:08, 1 September 2020 (UTC)
- Perhaps "determination method or standard (P459) = Template:According to EXIF data (Q18815612)" or "determination method or standard (P459) = Exchangeable image file format (Q196465)" qualifier. --Jarekt (talk) 13:22, 2 September 2020 (UTC)
- Currently we don't really do any sourcing so I'm just adding it. Multichill (talk) 19:50, 2 September 2020 (UTC)
- It is not the sourcing issue. For example I have about a year of photos taken with a camera that apparently had incorrect date set in it's settings, so I have bunch of photos from the first days of life of one of my kids and they all have a date off by over a year. if I upload one of those to commons and I do not provide the date they would be quite off. So "determination method or standard (P459) = Exchangeable image file format (Q196465)" qualifier would be to alert people that the date could be unreliable. Now that might be more of an issue with digital cameras an less of an issue with modern cell phones. --Jarekt (talk) 01:14, 3 September 2020 (UTC)
- Currently we don't really do any sourcing so I'm just adding it. Multichill (talk) 19:50, 2 September 2020 (UTC)
- Perhaps "determination method or standard (P459) = Template:According to EXIF data (Q18815612)" or "determination method or standard (P459) = Exchangeable image file format (Q196465)" qualifier. --Jarekt (talk) 13:22, 2 September 2020 (UTC)
Increase maximum precision for time datatype from day to second
[edit]See phab:T266407. Multichill (talk) 15:47, 25 October 2020 (UTC)
Adding general date values
[edit]One issue this modeling does not address is how to add dates when they are supplied by an institution without the same level of granular meaning as the properties being suggested here. For example, as aggregators, both the Europeana Data Model and the DPLA data model based on it have a general "date" field which can be used to encompass whatever type of date the source institution has added in a single common field (definition: "Use for a significant date in the life of the original analog or born digital object."). This is usually a creation date, but not necessarily, and might be a publication date, copyright date, broadcast date, performance date, etc. depending on the item and type of media. What property could we use to add such a date? Dominic (talk) 15:55, 13 December 2021 (UTC)
- Hi, we normally use inception (P571) throughout, i.e., for the most relevant date, which could be the publication date, copyright date, broadcast date, etc. depending on item and type of media. For example, normally for digital pictures we use the date the picture was taken, whereas we use for digital pictures of 2D representations of painting the date of the drawing not the date of the digital picture. --Schlurcher (talk) 20:55, 13 December 2021 (UTC)
- Great, I'm glad to hear that. I was afraid "inception" was only being applied strictly for date of creation. Thanks! Dominic (talk) 14:43, 14 December 2021 (UTC)
New recording date (P10135) property
[edit]This is just a notice that there is now a new recording date (P10135) the Wikidata community created. It can be used for videos, audio recordings, etc.
Maybe Commons might want to use it? Let me know what you think. Lectrician1 (talk) 21:28, 23 December 2021 (UTC)
- There are big benefits for using inception (P571) for the most relevant date troughout. I think consistency is more important here. The recording date (P10135)-Property may be an option for the parameter
photo date
in {{Art Photo}}, however imho it should not be used for the plain data parameter in any other if the {{Information}}-like templates. --Schlurcher (talk) 19:56, 25 December 2021 (UTC)- Using recording date sounds like a great idea. Especially for video and audio when recording is followed by sometimes heavy editing and then publishing makes it a more fluid process than photos. The recording can of course take place on a number of different dates before publishing which also speaks for inception not being great for this intermediate step. Ainali (talk) 18:32, 12 February 2022 (UTC)
Time data type "some value"
[edit]Does anyone know how to set a time (in inception (P571)) to "some value"? I can't figure out how. Tol (talk | contribs) @ 02:50, 28 December 2021 (UTC)
- So, I can do this through the API (diff), but I can't figure out a way to do it through the user interface. Tol (talk | contribs) @ 03:19, 28 December 2021 (UTC)
- @Tol: "some value" should only be used around the year 0 for inception (P571). In all other cases the precision should be adjusted to cover the possible range. See this example which does seem to trigger an old bug making it display "1999" at the moment. Multichill (talk) 14:00, 28 December 2021 (UTC)
- @Multichill: Ah; thank you — that makes sense! Tol (talk | contribs) @ 19:37, 28 December 2021 (UTC)
- @Tol: "some value" should only be used around the year 0 for inception (P571). In all other cases the precision should be adjusted to cover the possible range. See this example which does seem to trigger an old bug making it display "1999" at the moment. Multichill (talk) 14:00, 28 December 2021 (UTC)