The CRM giant’s Seq2SQL system is a deep neural network that translates natural language questions into corresponding SQL queries.
Salesforce wants to facilitate database queries for non-technical business users .
Salesforce’s Artificial Intelligence (AI) research team has developed a new system that promises to help business users communicate with databases without language skills, such as SQL, which is often used to query such systems.
The CRM giant’s Seq2SQL system, presented in an academic paper , is a deep neural network that translates natural language questions into corresponding SQL queries . Users write a question and the system returns results with the query to the appropriate database.
Salesforce has reported that its model is inspired by pointer networks instead of generating words from a fixed vocabulary – as does the model of attention from sequence to sequence.
By applying reinforcement learning, the Salesfoce system has been able to generate more accurate results than the attentional models from sequence to sequence. “Seq2SQL improved execution accuracy from 35.9% to 60.3% and accuracy from the logical form from 23.4% to 49.2%,” according to the company’s academic article.
The CRM giant is not the first company to attempt to transform the database query . The Business Intelligence Tableau firm, using ClearGraph’s technology, has a similar system to provide users with access and data analysis without any technical training.