Salesforce AI researchers developed a new system for helping business users communicate with databases without knowing languages, such as SQL. The new system, Seq2SQL, is a deep neural network, translates natural language questions to corresponding SQL queries.
According to Salesforce, the new model is inspired by pointer networks, which generate words from a fixed vocabulary like the attention sequence-to-sequence model, generates by selecting words from the input sequence.
Also, the new model generates more accurate results than attention sequence-to-sequence models. The company said, Seq2SQL improved execution accuracy from 35.9% to 60.3% and logical form accuracy from 23.4% to 49.2%.
The company also announced WikiSQL, an open-source dataset have more than 87,000 natural language questions, along with SQL queries, and SQL tables drawn from 26,000-plus HTML tables from Wikipedia. Salesforce is not a first company attempting to dumb down database querying.
Quepy, a Python framework transforms natural language questions into semantic database queries that can use with databases, such as DBpedia. Currently, Quepy supports for SPARQL and MQL query languages.
The data set built by extracting tables from Wikipedia and bringing them into a database engine. Salesforce then worked with people through Amazon Mechanical Turk to label the natural language queries for training a machine learning system.
Also, the data set helps other machine learning create systems like Seq2SQL. Other large public data set associated with major AI advances, like ImageNet, which uses to develop image recognition algorithms.
More information: [Salesforce]