Amazon Technologies, Inc.

Venkata Krishnan Ramamoorthy

Lokesh Joshi

Gregory Michael Hart

Samuel Scott Gigliotti

Christo Frank Devaraj

Christopher Allen Suver

Walter Manching Tseng

Ran Mokady

Robert Wai-Chi Chu

Zhaoqing Ma

Palanidaran Chidambaram

Canku Alp Calargun

Scott Southwood

Ilya Levner

Konstantinos Boulis

Shinobu Matsuzuka

Roman Yusufov

Prajwal Yadapadithaya

Gurbinder Gill

Ankur Jain

Hale Sostock

Pranap Sp

Kathirvel Thannasi

Patrick J Stammerjohn

Categorizing images may include training a first neural network to cluster a plurality of images to obtain a first image embedding space, wherein a vector representation is determined for each of the plurality of images based on the training, determining a vector norm value corresponding to each of the plurality of images based on the vector representation for each of the plurality of images, and identifying a first subset of the images for which a corresponding vector norm value satisfies a predetermined vector norm quality threshold. Then, a second neural network may be trained using the first subset of images to obtain a second image embedding space, and the second image embedding space may be used to categorize additional images.

Vector norm algorithmic subsystems for improving clustering solutions

Methods and systems for facilitating communications between shared electronic devices are described herein. In some embodiments, a group account may be assigned to a shared electronic device. The group account may include one or more user accounts, where individuals associated with those user accounts may interact with the shared electronic device, and also may form a part of the group account. When a message is sent from one shared electronic device to another personal device or shared electronic device, the message may be indicated as being sent from the group account, as if the shared electronic device corresponds to its own separate account. In some embodiments, speaker identification processing may be employed to determine a speaker of the message and, if the speaker is able to be identified, the message may be sent from the corresponding speaker's user account instead of the shared electronic device's corresponding group account.

Messaging from a shared device

This disclosure describes systems, methods, and computer-readable media related to contextual anchor points in content. In some embodiments, processed content may be received from a server. Contextual anchor parameters may be received from a user. A selection of an anchor target in the processed content may be received from a user. One or more contextual anchor candidates may be identified in the processed content based at least in part on the anchor target and the one or more contextual anchor parameters. The one or more contextual anchor candidates and at least a portion of the processed content may be displayed to the user. A selection of at least one contextual anchor from the one or more contextual anchor candidates may be received from the user. An association may be established between the anchor target and the contextual anchor. The processed content, anchor target, and the association may be transmitted.

Contextual anchor points in content

A user device receives a user input that causes an electronic publication to have a modified state. The user device generates data that, when used to render the electronic publication, causes the electronic publication to have the modified state. The user device transmits the data to a server.

Synchronizing electronic publications between user devices

Some examples include segmenting text of a content item to include a plurality of segments or words. For instance, a module for segmenting a content item using a context-based segmenter into a plurality of segments, identifying segment boundary hints stored in the content item, and adjusting segments of the plurality of segments based on the identified segment boundary hints. Some additional examples include inserting segment boundary hints into a content item. For instance a module that segments the content item using a first segmenter and a second segmenter and inserting segment boundary hints into the content item where the results of the first and second segmenter differ.

Determining word segment boundaries

Growing community of inventors

Tamil Nadu, India

Venkata Krishnan Ramamoorthy