Inventaire
Tale of a Wikidata-centered linked open contributive database on resources
Maxime Lathuilière
maxlath.eu - User:Maxlath - @maxlath
Berlin, 28 October 2017
find those slides at https://wiki.inventaire.io/wiki/Presentations-Slides
2 pillars
- book sharing web app
- open data on resources
1 mission
mapping resources with libre software and open knowledge (starting with books!)
The organization
-
3 years in a (figurative) garage
-
toward a non-profit structure (association loi 1901): a public service of private initiative
Project History in 5 seconds
- January 2015: inventaire.io is online!
- 2016: building the internal contributive open database to extend Wikidata
- 2017: documented API, NDJSON/TTL dumps, entities popularity scores
Projections
- 2017: SPARQL federation, reviews
- 2018: connecting to libraries & book shops inventories
- 2021: POTUS L. Lessig declares that all data on resources should be open, Amazon goes bankrupt
Web app tour
- inventories
- add books search/scanner/import
- users / groups / nearby
- transactions
- entities: authors, works, editions, series, subjects => popularity ranking
Where both pillars meet: the editor interface
Edit both Inventaire and Wikidata entities
ex: https://inventaire.io/entity/wd:Q3276730/edit
Data structure
URI-based Wikidata-mimicking data schema (labels, claims), but simplified and more strongly constrained to make edition easier by anyone
Entities map
RDF serialization
inv:1917e3d856a752b75e275743310dcc0c a wikibase:Item ;
rdfs:label "L'âge de la multitude"@fr ;
skos:prefLabel "L'âge de la multitude"@fr ;
wdt:P31 wd:Q571 ;
wdt:P50 inv:1917e3d856a752b75e275743310dc664,
wd:Q3132142 ;
wdt:P921 wd:Q173285,
wd:Q3908516 ;
wdt:P364 wd:Q150 .
If your books are not for sharing, but that you still want to use the data
data.inventaire.io
documenting the data model and everything
api.inventaire.io
documenting the HTTP/JSON API
dumps.inventaire.io
get raw NDJSON or Turtle dumps
query.inventaire.io
SPARQL endpoint federated to Wikidata (coming soon)
This is the data federation Wikidata needs:
- use Wikidata as the common vocabulary
- use domain-specific interfaces to get more eyes on the data, in a tailored, more constrained, user-friendly/proof way, ironing out data model issues (I'm looking at you WikiProject_Books)
Difficulties to build a web app on top of Wikidata and suggested work arounds
- typed search => subset search engine
- inconsistent data => keep a list of aliases (tolerant typing) and filter-out the rest
- expensive SPARQL queries take time => cache everything
- (controversial) guerrilla illustration: there are more images links in Wikidata to one can see at first look (social networks profile pics URLs, open graph, …)
The Future
graph-powered inventory browsers
Biggest difficulty: large graph parsing VS fast page load expectations
WIP: hidden feature
even more entities edition from all over the app
The Present
- pre-filled entity creation forms
The Future
- add books to a subject by clicking suggestions or by search => create a wdt:P921 claim, etc
The Future
book reviews, licensed CC-BY-SA by default, queryable from the API by Wikidata id
The Future
using Inventaire data to improve Wikidata:
- entity imports
- inferences. example: ISBN publisher prefix (P3035)
Sweets for the way
- sitelinks/external ids URI resolution
- .json
Your contribution is very welcome!
- improving the data
- discussing features: roadmap.inventaire.io
- code: git.inventaire.io AGPL
- translate: translate.inventaire.io
- help spread the word! (and don't forget your sticker!)