User talk:Multichill/Archives/2014/July

From Wikidata
Jump to navigation Jump to search

Rijksmuseum artworks data import

Hi Multichill, I read and noticed that you have yet begun to import data for artworks in the collection of Rijksmuseum (Q190804). Since I had a similar plan but don't have proper coding experience and atm not the time to change that I am very happy to see this!

User:Zolo
Jane023 (talk) 08:50, 30 May 2013 (UTC)
User:Vincent Steenberg
User:Kippelboy
User:Shonagon
Marsupium (talk) 13:46, 18 October 2013 (UTC)
GautierPoupeau (talk) 16:55, 9 January 2014 (UTC)
Multichill (talk) 19:13, 8 July 2014 (UTC)
Susannaanas (talk) 11:32, 12 August 2014 (UTC) I want to synchronize the handling of maps with this initiative
Mushroom (talk) 00:10, 24 August 2014 (UTC)
Jheald (talk) 17:09, 9 September 2014 (UTC)
Spinster (talk) 15:16, 12 September 2014 (UTC)
PKM (talk) 21:16, 8 October 2014 (UTC)
Vladimir Alexiev (talk) 17:12, 7 January 2015‎ (UTC)
Sic19 (talk) 21:12, 19 February 2016 (UTC)
Wittylama (talk) 13:13, 22 February 2017 (UTC)
Armineaghayan (talk) 08:40, 10 March 2017 (UTC)
Musedata102 (talk) 20:27, 26 November 2019 (UTC) Hannolans (talk) 18:36, 16 April 2017 (UTC)
User:Martingggg
Zeroth (talk) 02:21, 4 June 2018 (UTC)
User:7samurais
User:mrtngrsbch
User:Buccalon
Infopetal (talk) 17:54, 9 August 2019 (UTC)
Karinanw (talk) 16:38, 24 March 2020‎ (UTC)
Ahc84 (talk) 17:38, 26 August 2020 (UTC)
User:BeatrixBelibaste
Valeriummaximum
Bitofdust (talk) 22:52, 26 March 2021 (UTC)
Mathieu Kappler
Zblace (talk) 07:22, 24 December 2021 (UTC)
Oursana (talk) 13:16, 17 May 2022 (UTC)
Ham II (talk) 08:30, 25 January 2024 (UTC)

Notified participants of WikiProject Visual arts – they might be interested, too.

I assume that you have retrieved the data from the Europeana REST API and neither from Europeana Linked Open Data nor from the Rijksmuseum API (which provides some more structured data atm, the artworks size for example). Since Europeana covers more than 30 million items I think that your work will be the beginning of a quite big task for Wikidata, selecting of the part of the 30 million that shall be imported will be just one stone of it. I am very thankful that you published the code you used to used to import the data for artworks in the collection of the Mauritshuis (Q221092). But what do you think about using a repository e.g. on github for all the "Wikidata artwork data import" code? That would open possibilities to reuse and improve the code and to exchange improvements for me and others – Shonagon might be interested ;). I have yet two proposals to improve the code:

I hope the artworks on Wikidata will have a splendid future. Thanks another time for your recent work, --Marsupium (talk) 17:42, 8 July 2014 (UTC)

Yes I already saw this and it's great. I did notice a double, which will be inevitable I suppose for the more popular paintings (for which Commons has book scans, WGA versions, and now, Google art versions). I merged Q17319835 to Q2311238. Keep the bots running! This will be awesome. Jane023 (talk) 17:56, 8 July 2014 (UTC)
Hi Marsupium and Jane023, happy to see you like it!
I use the Europeana api. I looked at the LOD Europeana thingie and it didn't make me very happy, so I skipped that. All the important Rijksmuseum fields are available through the Europeana API so I didn't go to the Rijksmuseum API directly. The bot uses two steps:
  1. Find existing item or create an item which can be referenced later
  2. Expand the item with more useful things
I use the combination of collection (P195) and inventory number (P217) for the first step. Portrait of a Couple probably Isaac Massa and Beatrix van der Laen (Q2311238) didn't have this information so the bot couldn't find it and a duplicate was created.
The second step adds things like creator (P170). I guess depicts Iconclass notation (P1257) could be added to that
The current code is pretty custom and still sitting on my laptop. I'll probably check it into git at some point. I'm also working on a more general Pywikibot based data import tool. I consider these paintings and Rijksmonumenten imports test runs to learn how to build a more general tool. Next up is Frans Hals Museum (Q574961). Multichill (talk) 19:15, 8 July 2014 (UTC)
Hm. This will remain a problem with doubles then - I only stumbled on the marriage portrait because I am checking into all of the Frans Hals paintings. As you may have noticed by now, most bots just imported items to Wikidata from Wikipedia pages back in 2013 without any claims at all. Your guess of checking the collection and creator properties will probably not work unless someone has updated those by hand since then. I also found that you created and item for the portrait of Nicolaes Hasselaer, which I just adjusted to say that it depicts Nicolaes Hasselaer. For portraits I guess it may seem overkill to have a bio and a portrait item, but famous people will have had multiple portraits made, and I do think we want the entire collection of metadata for the world's top museums. Yay for FHM! It's much smaller I think, so it should be quick. Jane023 (talk) 19:36, 8 July 2014 (UTC)
(edit conflict) That seems real good, but the format used for the references does not match Help:Sources#Web_page. This should be fixed, if only because the recommended format matches Wikipedia's recommendation for citing sources, making it more suitable for Wikipedia transclusion.
Another question, is there a page listing artworks where the bot couldn't identify the item for the creator ? --Zolo (talk) 19:38, 8 July 2014 (UTC)
Before I start an import I first try to find existing items. I use https://tools.wmflabs.org/wikidata-todo/autolist.html for that. For Rijksmuseum I found about 30 existing items if I recall correctly. I looked up the inventory number for all of them.
I didn't even know we had Help:Sources#Web_page. I'll look into that.
You can find the paintings without creator here. Multichill (talk) 20:08, 8 July 2014 (UTC)
We are obviously all working on similar problems …
  1. Thinking about the three existing APIs I came to the same solutions as you. I think the Europeana REST API is a good choice for the moment. I am looking forward to the publication of your code on the Wikimedia git!
  2. From this version of Portrait of a Couple probably Isaac Massa and Beatrix van der Laen (Q2311238) it was impossible to identify the artwork for a bot (and even a human staying on Wikidata, not using Wikipedia). So the matching problem has to be solved in general. For that we have to strengthen the aim to reduce the number of artwork items that do not make the artworks unambiguously identifiable. (I is probably even appropriate to point out a criterion for this on WD:N.) Though it is a difficult problem to make all artwork items unambiguously identifiable. The combination of collection (P195) and inventory number (P217) is one and probably one of the most successful approaches (cf. also the current discussion WD:Project Chat#Gallery/ museum accession numbers). The number of just painting (Q3305213)(-subclass)-instance-items without at least a collection (P195) or a location (P276) statement is currently 4077 according to autolist and this is only a subset of the problem. It has to become 0.
  3. I actually thought of the processing of depicts Iconclass notation (P1257) you have described.
  4. The more general Pywikibot based data import tool you are talking about, is something I have thought about, too. I have in mind a Pywikibot script simply importing ready data in the Wikidata JSON format. The creation of the input files for this script and the matching algorithm could then be separated code and separated between some users.
Cheers, --Marsupium (talk) 15:25, 9 July 2014 (UTC)

Re about creators: actually they can be added pretty quickly with autolist2. But if done this way, they will be unsourced, so it would be better if the bot could add it directly. That leads me to yet another question: is there anything we can do to help the bot identify artists ? Adriaen van Ostade seems to be identified by RM0001.PEOPLE.1547 and I imagine we could add it somewhere in Adriaen van Ostade (Q352438)? But I can't find any documentation about this format in Europeana, so I am not sure what to do with it.--Zolo (talk) 16:01, 9 July 2014 (UTC)

Importing coord from itwiki

Hi! When you have time, there is an open bot request that may interests you. Thanks! --AlessioMela (talk) 20:32, 10 July 2014 (UTC)

Rude tone?

ist das das Einzige, was du zu sagen hast, wenn jemand schon fast verzweifelt versucht, den hier stattfindenden Schwachsinn zu begrenzen?

Wenn du dich an amerikanische Diskussionsverhältnisse anpassen willst (freundliches Lächeln mit der Axt in der Hand, bereit zum Zuschlagen), dann mach es irgendwo. Aber Wikidata ist kein amerikanisches Projekt. Bitte keine weiteren solcher Beiträge, entweder du gehst inhaltlich darauf ein oder du verzichtest darauf.--Hubertl (talk) 22:06, 14 July 2014 (UTC)

Retrieved from "https://www.wikidata.org/w/index.php?title=User_talk:Multichill/Archives/2014/July&oldid=151764791"