Mathias Dillen

Frau (@vonderleyen) sollte punkto #wolf Regulierung über der Sache stehen. 7 tote Wölfe für einen elusiven GW950m (Ponykiller), Planung de Änderung Schutzstatus des Wolfs in der @EU_Commission und keinen adäquaten Schutz ihrer Pferde zeugen nicht davon. theguardian.com/environment/20…

@emhaston @rdmpage @JSTORPlants @GBIF @tdwg It'll be up for public review the next round. github.com/tdwg/dwc/issue…

“Where are the plant type specimens? Mapping @JSTORPlants to @GBIF” iphylo.blogspot.com/2023/10/where-… #blogpost I attempted to match 1.3 million plant type specimens in @JSTORPlants to the record for the same specimen in @GBIF. Got 903,945 (67%).

@rdmpage @GBIF @JSTORPlants GBIF is actively updated, JSTOR was often dump-populated 10+ years ago. So it's not that surprising. Given how difficult it is to access JSTOR content, it seems more productive to focus on GBIF data and match with taxon name databases for type links.

@MathiasDillen @GBIF @JSTORPlants But it just seems bonkers that there’s little alignment between these two major digitisation efforts (@GBIF and @JSTORPlants)

Working on linking @JSTORPlants type specimens to corresponding occurrences in @GBIF, this is not nearly as easy as it should be. Matching identifiers is hard when people insist on mangling them, remixing them, or just deleting them. #PIDs

@rdmpage @GBIF @JSTORPlants It'll be mainly differences in file compression and dimensions. Rescanning or other changes post publication should be rare. I daresay that the hash method would work for the majority of specimen images. And if not, it would be intriguing to find out exactly why.

@MathiasDillen @GBIF @JSTORPlants So we’d need to test for image identity (same file), derived image (same image different size), and same thing but different image (although this case seems rare).

@rdmpage @GBIF @JSTORPlants Reverse image search as a service is not that easy. Easiest could be if GBIF implemented a hash archive for the media references people publish? Wouldn't work for derivative images, but would make it easier to go through the broken id cleanup routine.

@GBIF @JSTORPlants Of course, this image search feature in @GBIF doesn’t actually exist, so there’s that. But how hard can it be to index images and make them searchable à la images.google.com 🙈

@BionomiaTrack @frictionlessd8a We're planning some work on roundtripping from Bionomia in @Bicikl_H2020, so definitely interested in contributing.

How are you using crowdsourcing platforms? And how should it work in the future? We want to hear your voice!!! Please fill out our survey 👃 forms.gle/mQcB41NV9ddi5K… @DoeDatbe @BGM_coll_res @DiSSCoEU @eurotaxonomy @tdwg @DARIAHeu @LifeWatchERIC @LifeWatchVLIZ

@RBGE_Plant_Rec @Pensoft @BGM_coll_res @GBIF @tdwg @DiSSCoEU @Bicikl_H2020 github.com/agentschapplan…

🔥LIVE from #TDWG2022: @MathiasDillen (@BGM_coll_res) on what is Minimum information about a digital #specimen (MIDS). He goes on to present the Rshiny app allowing for easy calculation of MIDS to be used in @GBIF datasets. @tdwg @DiSSCoEU @Bicikl_H2020

Our entry to the 2022 Ebbe Nielsen challenge: Get the digitization level for your published #GBIF specimens according to the current (and future) specifications of the MIDS standard. youtube.com/watch?v=117547… @cabbageleek @emhaston @HuyPieter @AlexHardisty @CatDigsBio @tdwg @gbif

@andrawaag @GBIF @rdmpage Calling it occurrence ID is a bit misleading, as these records also have a dwc:occurrenceID that is completely different. These IDs should be more persistent, but they can also change.

Yesterday we proposed the #wikidata property for the @GBIF occurrence ID. I now learn from @rdmpage that those identifiers are not persistent and even frequently change. I guess we should retract the proposal right? property: wikidata.org/wiki/Wikidata:…

@dpsSpiders @GarretsonAlexis @rdmpage @GBIF @BionomiaTrack @plazi_treat @Bicikl_H2020 @myrmoteras @wikidata @dnnyboy That, but it's also not easy to systematically extract individual specimen citations from literature. They can be anywhere in the documents, in any format.

Are there plans through @GBIF to track literature usage down to the specimen level? The dataset, publisher, and download citation counters are so awesome, it would be awesome to have that built-in at the specimen level (I know @BionomiaTrack is doing great stuff here too!)

@rdmpage @dpsSpiders @GarretsonAlexis @GBIF @BionomiaTrack @plazi_treat @Bicikl_H2020 @myrmoteras @wikidata We've tried this and it works to some extent. A common form of citation includes a taxon name, a coll. code and/or a person (sur)name. Clustering along these lines, w some taxon id/rank wrangling, code regex and fuzzy surnames goes a long way.

@dpsSpiders @MathiasDillen @GarretsonAlexis @GBIF @BionomiaTrack @plazi_treat @Bicikl_H2020 @myrmoteras @wikidata I often think that going in the reverse direction is likely to be powerful, that is, generate the strings you'd expect to see cited (say, a combination of locality, date, collector code, etc.) from @GBIF data then search text for those. See also iphylo.blogspot.com/2021/05/findin…

On the way to liberate the riches of data and links about specimen, genes hidden in publications, especially tables in #15. Right now an almost complete, frustrating disconnect. plazi.org/posts/2021/11/… #BioHackEU21 #BioHackEU21 @Bicikl_H2020 @ELIXIREurope

Check out my poster at #TDWG2021, made possible by @DisscoFlanders and @Microsoft #AIforEarth Planetary Computer Can we estimate the completeness of collections with stats and also compare them to the whole of @GBIF ? short answer: yes 🔗doi.org/10.3897/biss.5… @BGM_coll_res

Oooh, Now! It's #biodiveritydata #knowledgegraphs @TDWG #TDWG2021 with @andrawaag @baskaufs @mdmtrv @MathiasDillen and Elie Mario Saliba

@lj_garcia @ZB_MED @ELIXIREurope @biohackathon Can descriptions/abstracts be up to 200 or 400 words? Guidelines on the BH site and on easychair don't agree.

Did you submit already your hacking project idea? We @ZB_MED did 🙂 Do not miss this chance, submissions open until 1st April biohackathon-europe.org #BioHackEU21 @ELIXIREurope @biohackathon

@dpsSpiders Data files should be watermarked somehow to indicate if they were ever edited with spreadsheet software. One thing that may help is to always call the first column in a csv file ID. Excel will neatly refuse to open it. journeybytes.com/fix-csv-sylk-e…