“The big disruption here is that we’ve made it simple and commercially viable to build a custom AI model for your business intelligence team with limited labeling data provided by the company,” said Peter Relan, Chairman and CEO of Got It.
BURLINGAME, Calif. (PRWEB) June 19, 2019
Got It, the first NLP+Human-in-the-Loop platform embeddable in SaaS and PaaS products, announced today the first Google BERT-based Natural Language Understanding (NLU) Transfer Learning Service for companies with data warehouses. TransferML for BigQuery is pre-trained on Google BigQuery with CRM data and Google Analytics 360 data, enabling each customer to build a custom, business-specific AI model and deploy it within a week. Usage-based pricing dramatically reduces BI costs for any organization while extending the capability for data-driven insights to every knowledge worker in every department. Details about TransferML for BigQuery will be revealed at the Re-Work Deep Learning Summit in San Francisco on June 20-21.
“The big disruption here is that we’ve made it simple and commercially viable to build a custom AI model for your business intelligence team with limited labeling data provided by the company,” said Peter Relan, Chairman and CEO of Got It. “The power of BERT-based transfer learning and our pre-trained model on Salesforce and GA 360 data is that customers can start saving big dollars, or drastically reduce BI backlog, in one week. I’m delighted that we can fulfill Andrew Ng’s previous prediction that transfer learning would become the next big driver for commercial adoption of machine learning.”
Unlike the inherently limited, keyword-based, question-answering analytics tools available today, TransferML for BigQuery leverages an advanced neural architecture for real Natural Language Understanding. This has been made possible through two recent technical breakthroughs:
1. Google BERT, a pre-trained language encoder, fine-tuned to write simple SQL queries using the public WikiSQL dataset in research by scientists at NAVER Corporation.
2. Got It’s proprietary process to further tune the model for specific business schemas and more complex queries.
In addition, the rapid training and deployment of custom models for each company’s BI team is made possible by Got It’s managed marketplace of SQL experts.
“Unlike Mechanical Turk or Upwork, Got It leverages an on-demand, fully managed marketplace to bring vetted analytics experts to help with labeling within seconds, plus proprietary operational processes for labeling data, enabling BI teams to build custom business-specific models rapidly with high accuracy rates,” said Utkarsh Mittal, Got It AI Operations Lead on TransferML for BigQuery.
TransferML for BigQuery is currently being offered in pilot mode to a limited number of Google BigQuery customers. Production APIs combined with the labeling service will be released later this year and support for Azure SQL Data Warehouse, RedShift, Oracle, and additional SaaS software schemas will be added over time.
TransferML for BigQuery is also the underlying service that powers the recently announced Got It Querychat product: a full application and workflow for BI teams and end-users to deploy immediately and then use the custom AI model generated automatically for their use case. Querychat becomes available in July 2019, followed by TransferML for BigQuery in October 2019.
About Got It
Got It has developed the first NLU+Human-in-the-Loop platform that can be embedded in any SaaS or PaaS product to make product use experiences better, smarter, and more efficient. Users frequently get stuck on how to use a SaaS or PaaS product feature or capability. An AI solution is available within a second, backed up by a vetted human expert available within 30 seconds to assist if needed. All users have to do is chat in plain English, and Got it guarantees that they will get the solutions they need. We're boosting the productivity of knowledge workers who use SaaS or PaaS products from Google, Microsoft and Oracle. For more information go to https://www.got-it.ai/