公开(公告)号：US11288513B1

公开(公告)日：2022-03-29

申请号：US16675383

申请日：2019-11-06

Applicant: Amazon Technologies, Inc.

Inventor： Chintan Desai ,  Tarun Venkatesh Ravva ,  Yen-Yi Lai ,  Prathyusha Moparthi ,  Tong Wang ,  Anthony Payauys

IPC: G06K9/00 ,  G06K9/72 ,  G06K9/46 ,  G06F40/30 ,  G10L15/22 ,  G10L15/18 ,  G10L13/00

Abstract: Described herein is a system for predictive image analysis to provide a response to a user input. To reduce a system response time when responding to user inputs that relate to image data analysis, the system processes data relating to past user inputs to predict when a user may request information relating to image data. Using pattern data (determined by the system), the system determines that a trigger to initiate image analysis occurred. The trigger may indicate that the system is expected to receive a user input relating to image data. The system processes image data, in advance of receiving the user input, to determine and store feature data that can be used to generate an output responsive to a user input. Then when the user input is received, the system sends output data (generated using the determined feature data) responsive to the user input.

公开(公告)号：US20230074681A1

公开(公告)日：2023-03-09

申请号：US17866965

申请日：2022-07-18

Applicant: Amazon Technologies, Inc.

Inventor： Han Wang ,  Tong Wang ,  Yue Liu ,  Ashish Kumar Agrawal

IPC: G10L15/18 ,  G10L15/26

Abstract: Techniques for processing complex natural language inputs are described. A complex natural language input may be semantically tagged and parsed to identify individual clauses in the complex natural language input. An execution graph may be generated to represent the clauses and their dependencies. Nodes of the execution graph may be processed using NLU processing and/or a knowledge graph or other information storage and retrieval techniques, and results of such processing may be used to update clause variables with specific entities in the execution graph.