Amazon Technologies, Inc.

Jingxiang Chen

Xiang Hao

Vernon Germano

Muhammad Raffay Hamid

Ahmed Aly Saad Ahmed

Mohamed Kamal Omar

Yash Chaturvedi

Steven James Cox

Jared Kramer

Alexander Ratnikov

Prasanth Saraswatula

Simi Wang

Hebatallah Elfardy

Andrea Kahn

Lakshay Sharma

Systems, devices, and methods are provided for determining contextually relevant user-based product recommendations based on scene information. In at least one embodiment, techniques described herein may be used to determine, using a first machine-learning model, first information associated with a first object within a first image of digital content, determine, using a second machine-learning model, similarity scores between the first object and a first plurality of products of an online purchasing system, detect, in association with the first image of the digital content, performance of a first computer-based action by a user, determine, using a third machine-learning model and based on contextual data of the user, one or more affinity scores for the user, select a first product based on the one or more affinity scores, and present a recommendation to the user to perform a second computer-based action in association with the first product.

Contextually relevant user-based product recommendations based on scene information

Devices, systems, and methods are provided for context-based abusive language detection and responses. A method may include identifying text associated with first video content, and determining that a first word in the text matches a first keyword indicative of abusive language. The method may include determining a first label associated with the first word, the first label indicating that the first word is ambiguous. The method may include identifying a first sentence of the text, the first sentence including the first word. The method may include determining first and second context of the first word and the first sentence. The method may include determining, based on the first and second context, using a machine learning model, a second label associated with the first sentence, the second label indicating a probability that the first sentence includes abusive language. The method may include generating second video content for presentation.

Context-based abusive language detection and response for media

Techniques for temporal localization of mature content in long-form videos using only video-level labels are described. According to some embodiments, computer-implemented method includes receiving a request to train a machine learning model on a training video file comprising at least one mature content label, training the machine learning model to generate a feature vector for each of a plurality of video frames of the training video file, generate a plurality of frame-level mature content classification scores of the training video file from the feature vectors of the training video file, and generate a video-level mature content classification score of the training video file from the plurality of frame-level mature content classification scores for the training video file based at least in part on the at least one mature content label of the training video file, receiving a request for an input video file, generating, by the machine learning model in response to the request, a feature vector for each of a plurality of video frames of the input video file, a plurality of frame-level mature content classification scores of the input video file from the feature vectors of the input video file, and a video-level mature content classification score of the input video file from the plurality of frame-level mature content classification scores for the input video file, and transmitting the plurality of frame-level mature content classification scores of the input video file or the video-level mature content classification score of the input video file to a client application or to a storage location.

Temporal localization of mature content in long-form videos using only video-level labels

A pipeline is provided for management of a pool of chat message templates for an automated dialog system. The pool of chat messages may be managed using machine learning-based clustering and feedback-based modifications. A set of chat messages may be analyzed using a machine learning model to generate different clusters of messages that are semantically related. Representative messages may be selected from each cluster and used in chat sessions according to the semantic context of the chat sessions. Based on feedback obtained during the chat sessions, metrics generated based on the feedback, and/or other data, modifications may be made to the clusters and/or the representative messages to improve the performance of the automated dialog system.

Architecture for dynamic management of dialog message templates

Growing community of inventors

Bellevue, WA, United States of America

Jingxiang Chen