Fabrizio Silvestri: We just released on arXiv (https://arxiv.org/abs/2010.06973) our new paper on "Neural Databases".
We presents NeuralDB, a DB system with no pre-defined schema, in which updates & queries are given in natural language.
@j6mes, @myazdani999, @marzieh_saeidi, @riedelcastro, and Alon Halevi
4 replies, 198 likes
Sebastian Riedel: With Neural Theorem Provers we always tried to inject neural components into symbolic architectures. I like to think of the NeuralDB as the opposite: we inject a few symbolic operators into the neural architecture. And this works well! https://t.co/bOM4XwhODg
0 replies, 72 likes
Pasquale Minervini: Love this! https://t.co/gnbCXhK32R
1 replies, 19 likes
Daisuke Okanohara: Neural Databases is a database w/o schema; data are sentences with facts, queries are also a sentence specifying select/project/join operations. transformer model pretrained w/ LM computes the result. scaling up neural DB to larger databases is the issue. https://arxiv.org/abs/2010.06973
0 replies, 4 likes
HotComputerScience: Most popular computer science paper of the day:
0 replies, 3 likes
Joy Gao: Thought-provoking paper.
For the most part I’ve thought of ML’s role in database evolution as indexing.
But this paper proves that it could go beyond that, storage and querying could look quite different in the future too.
0 replies, 2 likes
Found on Oct 16 2020 at https://arxiv.org/pdf/2010.06973.pdf