Open Data Project Used to Enable AI Applications Is Now Queried More Than 250,000 per Day
Luminoso, the natural language company that provides AI-powered analysis of unstructured data, has announced new features and adoption figures for ConceptNet, an open data semantic network created by Luminoso’s co-founders.
ConceptNet is designed to help computers understand the meanings of the words that people use. ConceptNet originated from MIT Media Lab’s Open Mind Common Sense project, and is now used in AI applications around the world. ConceptNet’s API is queried over 250,000 times per day from more than 1,000 unique IPs. The open data semantic network is cited in more than 500 AI papers in Google Scholar.
“In an era where many still build custom ontologies for their machine learning applications, a growing number of people have discovered firsthand that computers still gain more common-sense understanding with ConceptNet,” said Robyn Speer, Luminoso’s chief science officer who also leads the development of ConceptNet 5. “Every update we make to ConceptNet advances machine learning applications all over the world, in dozens of different languages.”
ConceptNet 5.7 features:
- Database updates: ConceptNet is now powered by PostgreSQL 10. Previously, ConceptNet’s architects had to write carefully-tuned SQL queries so that the server could quickly respond to ConceptNet API queries. The new database structure offers a more efficient way to do every kind of ConceptNet query, without the need for manual tuning.
- New Japanese knowledge: ConceptNet 5.7 includes 18,747 new facts in Japanese, covering common-sense relations that are hard to collect data for, such as where certain objects or subjects are located, and what objects are used for.
- Word senses: While ConceptNet previously offered limited support for word senses, defined as words that have multiple meanings, ConceptNet now provides more details about word senses.