公开(公告)号：US12175966B1

公开(公告)日：2024-12-24

申请号：US17361003

申请日：2021-06-28

Applicant: Amazon Technologies, Inc.

Inventor： Yi-An Lai ,  Yi Zhang ,  Roger Scott Jenke ,  Meghana Puvvadi ,  Shang-Wen Daniel Li ,  Peng Zhang ,  Jason P. Krone ,  Garima Lalwani ,  Niranjhana Nayar ,  Kartik Natarajan

IPC: G10L15/22 ,  G06N5/04 ,  G06N20/00 ,  G10L15/06 ,  G10L15/18 ,  H04L51/02 ,  G10L15/16

Abstract: Techniques for updating a machine learning model based on user interactions are described. In particular, in some examples, user interactions with a chatbot provide aspects of a data set to be used to train or fine-tune a ML model. In some examples, this is accomplished by collecting data from a first plurality of interactions with a machine learning (ML) model; generating a variant of the ML model using the collected data by: filtering the collected data to create a first data set, training the ML model based on the first data set to generate an adapted ML model, and fine-tuning the adapted ML model on a second data set, different than the first data set to generate the variant of the ML model.

公开(公告)号：US11580968B1

公开(公告)日：2023-02-14

申请号：US16455165

申请日：2019-06-27

Applicant: Amazon Technologies, Inc.

Inventor： Arshit Gupta ,  Peng Zhang ,  Rashmi Gangadharaiah ,  Garima Lalwani ,  Roger Scott Jenke ,  Hassan Sawaf ,  Mona Diab ,  Katrin Kirchhoff ,  Adel A. Youssef ,  Kalpesh N. Sutaria

IPC: G10L15/08 ,  G10L15/16 ,  G10L15/22 ,  H04L67/02 ,  G10L15/18 ,  G10L15/02

Abstract: Techniques are described for a contextual natural language understanding (cNLU) framework that is able to incorporate contextual signals of variable history length to perform joint intent classification (IC) and slot labeling (SL) tasks. A user utterance provided by a user within a multi-turn chat dialog between the user and a conversational agent is received. The user utterance and contextual information associated with one or more previous turns of the multi-turn chat dialog is provided to a machine learning (ML) model. An intent classification and one or more slot labels for the user utterance are then obtained from the ML model. The cNLU framework described herein thus uses, in addition to a current utterance itself, various contextual signals as input to a model to generate IC and SL predictions for each utterance of a multi-turn chat dialog.

公开(公告)号：US11562735B1

公开(公告)日：2023-01-24

申请号：US16836130

申请日：2020-03-31

Applicant: Amazon Technologies, Inc.

Inventor： Arshit Gupta ,  Julian E. S. Salazar ,  Peng Zhang ,  Katrin Kirchhoff ,  Yi Zhang

IPC: G10L15/18 ,  G10L15/197 ,  G10L15/26

Abstract: A spoken language understanding (SLU) system may include an automatic speech recognizer (ASR), an audio feature extractor, an optional synchronizer and a language understanding module. The ASR may produce a first set of input data representing transcripts of utterances. The audio feature extractor may produce a second set of input data representing audio features of the utterances, in particular, non-transcript specific characteristics of the speaker in one or more portions the utterances. The two sets of input data may be provided for the language understanding module to predict intents and slot labels for the utterances. The SLU system may use the optional synchronizer to align the two sets of input data before providing them to the language understanding module.

公开(公告)号：US10904669B1

公开(公告)日：2021-01-26

申请号：US16712299

申请日：2019-12-12

Applicant: AMAZON TECHNOLOGIES, INC.

Inventor： Shahram Talakoub ,  Jianchun Dong ,  Peng Zhang ,  Yuwen Su ,  Xuan Zhong

IPC: H04R1/32 ,  H04R5/04 ,  H04R3/04 ,  H03G5/16 ,  H04S7/00 ,  G02B27/01 ,  H04R1/02 ,  H04R1/10 ,  H04R1/20 ,  H04R3/12

Abstract: A head-mounted wearable device (HMWD) provides audio output from a first speaker that is driven with a first signal and a second speaker that is driven with a second signal. Based on a volume level setting, an equalization profile and inversion frequency are determined. The equalization profile selectively amplifies or attenuates particular frequencies or ranges of frequencies. Those frequencies in the second signal that are above the inversion frequency have their amplitude inverted, relative to the first signal. When driven by the first signal and the second signal, the first speaker and the second speaker operate as acoustic dipoles below the inversion frequency and acoustic quadrupoles above the inversion frequency. Sound from the first and second speakers with frequencies above the inversion frequency exhibits destructive interference. As a result, the user wearing the HMWD is able hear audio output while audio amplitude perceived by the bystanders is significantly reduced.