公开(公告)号：US20220121685A1

公开(公告)日：2022-04-21

申请号：US17072597

申请日：2020-10-16

Applicant: salesforce.com, inc.

Inventor： Zuye Zheng ,  Scott Musser ,  Raghav Ravichandran ,  Asra Yousuf ,  Meghana Ashok ,  Andrew Ouyang ,  Aditya Kochhar

IPC: G06F16/28 ,  G06F16/242 ,  G06F16/9032 ,  G06K9/62 ,  G06N20/20

Abstract: Methods, systems, and devices for generating a query using training observations are described. According to the techniques described herein, a device (e.g., an application server) may receive a set of queries including a set of fields in a tenant-specific dataset associated with the query. The device may generate a set of training observations for the queries based on the set of fields. The device then trains a first machine learning model to determine grouping hierarchies and a second machine learning model to determine aggregation predictions. The device then builds a combined machine learning model based on the determined grouping hierarchies and the aggregation predictions. According to techniques described herein, the device uses the determined grouping hierarchies and the aggregation predictions to rank a set of suggested queries determined in response to an input query and selects a suggested query for display based on the ranking.