[2010.06973] Neural Databasesopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

In recent years, neural networks have shown impressive performance gains on long-standing AI problems, and in particular, answering queries from natural language text. These advances raise the question of whether they can be extended to a point where we can relax the fundamental assumption of database management, namely, that our data is represented as fields of a pre-defined schema. This paper presents a first step in answering that question. We describe NeuralDB, a database system with no pre-defined schema, in which updates and queries are given in natural language. We develop query processing techniques that build on the primitives offered by the state of the art Natural Language Processing methods. We begin by demonstrating that at the core, recent NLP transformers, powered by pre-trained language models, can answer select-project-join queries if they are given the exact set of relevant facts. However, they cannot scale to non-trivial databases and cannot perform aggregation q

21 mentions: @fabreetseo@icoxfog417@hillbig@hillbig@jreuben1@joemccann@mickeymcmanus@waynerad
Date: 2020/10/17 18:53

Referring Tweets

@hillbig 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. t.co/PliOHm4OKZ
@icoxfog417 自然言語でデータを格納し自然言語で問い合わせを行うNeuralDBの提案。factを含む文をデータ、factを問う文をクエリとして構成する(複数factの連結が必要な質問=joinも可)。通常知識グラフの構築からSPARQLで問い合わせるが、Transformer(T5)を使用し事前関係不要にしている t.co/d3h21Tz62t
@jreuben1 NeuralDB t.co/wuSA02aK1v T5 Transformer runs multiple Neural SPJ operators in parallel, fed to an aggregation operator (decoder fusing) -> answer queries over thousands of sentences with very high accuracy t.co/La0l5cdv1T
@hillbig Neural Databasesはスキーマ不要で事実を含む自然文をそのままDBに登録、Select/Project/Join/部分的にAggregation操作が自然文で書かれたクエリに対し、DB登録文(サポート集合に絞る場合も)とクエリをつなげた入力をTransformerで処理し結果を返す。スケール化が課題t.co/PliOHm4OKZ
@joemccann This feels like the natural evolution of neural networks research is to extend where the data is being stored itself -- neural databases. t.co/seXTL5iVZg
@fabreetseo We just released on arXiv (t.co/LY05D1aOLr) 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
@3manuek "...powered by pre-trained language models, can answer select-project-join queries if they are given the **exact set of relevant facts.**" t.co/obQtCbKo2E Queries be like: t.co/Q1P69DyJWs
@waynerad NeuralDB is a database system in which updates and queries are given in natural language. Who is Sheryl’s husband? (Lookup) -> Nicholas t.co/qP6coyor0W

Related Entries

Read more [1911.06311] Sato: Contextual Semantic Type Detection in Tablescontact arXivarXiv Twitter
0 users, 1 mentions 2020/03/10 08:20
Read more [2004.11362] Supervised Contrastive Learningopen searchopen navigation menucontact arXivarXiv Twitte...
1 users, 15 mentions 2020/04/24 05:24
Read more [2004.13912] Neural Additive Models: Interpretable Machine Learning with Neural Netsopen searchopen ...
0 users, 10 mentions 2020/04/30 14:21
Read more [2007.03898] NVAE: A Deep Hierarchical Variational Autoencoderopen searchopen navigation menucontact...
0 users, 14 mentions 2020/07/10 00:53
Read more Semi-Supervised Learning of Protein Secondary Structure from Single Sequences | bioRxiv
0 users, 4 mentions 2020/07/20 03:50