Freebase: a fact base you could write to
On 9 June 2008, at the SIGMOD conference, the Metaweb Technologies team described Freebase: a knowledge base of more than 125 million tuples, more than 4,000 types and more than 7,000 properties. Read and write access was open to anyone over an HTTP API with its own query language, MQL.
Why it matters
Until then a large fact base either belonged to one organisation or was a snapshot posted for download. Freebase did a third thing: public write access. An outsider could not only read a fact but add one, and the next query would see it. Four years later Google's search sat on this base.
The authors were Kurt Bollacker, Colin Evans, Praveen Paritosh, Tim Sturge and Jamie Taylor. The paper was a demonstration, pages 1247-1250. The figures from its abstract: more than 125,000,000 tuples, more than 4000 types, more than 7000 properties. The same abstract states that public read/write access is allowed through an HTTP-based graph-query API using the Metaweb Query Language. That the base was open long before this paper has an independent trace: on 9 March 2007 Creative Commons wrote about Freebase as Metaweb's project adopting the CC Attribution licence, quoting it: "All data is licensed Creative Commons Attribution. We only ask that you link back to us." What this record does not claim. It names no launch day in 2007: none of the sources read gives one, and the attribution to the O'Reilly ETech conference did not hold up. Nor does the record describe the internals - neither compound value types nor the graphd store: the full text is served only as a download and was not read, and the abstract says nothing of them.