公开(公告)号：US11929070B1

公开(公告)日：2024-03-12

申请号：US17461124

申请日：2021-08-30

Applicant: Amazon Technologies, Inc.

Inventor： Ruhi Sarikaya ,  Zheng Du ,  Xiaohu Liu ,  Kai Liu ,  Sriharsha Venkata Chintalapati ,  Chenlei Guo ,  Hung Tuan Pham ,  Joe Pemberton ,  Zhenyu Yao ,  Bigyan Rajbhandari

IPC: G10L15/22 ,  G06N20/20 ,  G10L15/02 ,  G10L15/06

CPC classification number: G10L15/22 ,  G06N20/20 ,  G10L15/02 ,  G10L15/063 ,  G10L2015/225

Abstract: Techniques for performing centralized unsuperivised learning in a multi-domain system are described. A user may request labeled data for an ML task, where the request includes a prompt for obtaining relevant explicit user feedback. The system may use the prompt to collect explicit user feedback for relevant runtime user inputs. After a duration of time (in the user's request for labeled data) has elapsed, the system determines whether collected user feedback indicates processing of the user input was defective and, if so, determines a cause of the defective processing. The system then uses one or more label generators to generate labeled data using the collected user feedback, whether the processing was defective, and the determined defect cause.

公开(公告)号：US09971751B1

公开(公告)日：2018-05-15

申请号：US14610716

申请日：2015-01-30

Applicant: Amazon Technologies, Inc.

Inventor： Vanessa Y. Sun ,  Feras Karablieh ,  Xiaohu Liu ,  Ajay M. Desai

IPC: G06F17/00 ,  G06F17/22 ,  G06F17/24

CPC classification number: G06F17/2288 ,  G06F17/2247 ,  G06F17/241 ,  G06F17/243 ,  G06F17/248 ,  G06F17/30876

Abstract: Multiple versions of a service can be supported by taking advantage of the fact that version information can be included in any received request. Multiple templates may be stored, with each template being associated with a particular version of the service. There may also be multiple templates associated with each version of a service for various response types. Using the version information, an appropriate template can be determined up front to be used for the response. An instance of the template can be filled in using appropriate data sources, as may be specified using metadata in the template itself. Once the instance of the template is filled, the response can be finalized and sent.

公开(公告)号：US11544504B1

公开(公告)日：2023-01-03

申请号：US17022883

申请日：2020-09-16

Applicant: Amazon Technologies, Inc.

Inventor： Xing Fan ,  Hung Tuan Pham ,  Chenlei Guo ,  Xiaohu Liu ,  Shuting Tang

IPC: G06K9/62 ,  G06F16/9032 ,  G06F40/35

Abstract: Techniques for determining an intent of a subsequent user input in a dialog are described. The system processes historic interaction data that is structured based on natural language understanding (NLU) hypotheses, with each NLU hypothesis being associated with one or more past user inputs received by the system, one or more sample inputs, and one or more past system responses. Based on processing of the historic interaction data and dialog data of previous turns of the dialog, the system determines candidate intents for the subsequent turn of the dialog. The system also uses context data to determine the candidate intents.