Amazon Technologies, Inc.

A9.com, Inc.

Alexandru I Indrei

Brent Mills

Blair Livingstone Hotchkies

Daniel R Parshall

Jason P Patrikios

Sarah Graham Williams

Sikandar Saifullah

John Michael Nilles

Sean Phillip Forde

Michel Leonard Goldstein

Michael Lee Brundage

Walter Manching Tseng

Paul Daniel Jaye

Sachin Birendra Singh

Max Lee Kanter

Pongsakorn Teeraparpwong

David Charles Couvrette

Carl N Meister

Matthew W Amacker

Daniel Eric Rose

Dan Catalin Teodorescu

Anil A Sewani

Brent Robert Mills

Rui Chen

Amber Autrey Taylor

David James Roy Erdmann

Daniel Thomas Mellott

Gerard Gjonej

Raghu Deep Gadde

Jennica Jane Pounds

David Aaron Lichterman

Aleix Margarit Martinez

Josh Sternberg

Anqi Liang

Mengyun Lv

Qianli Feng

Joyce Huan Fu

Lenworth Richard Rose

Ethan Alexander Smith

Xiaopeng Zhang

Akshad Viswanathan

Saeed Salahi

Reece Dano

Sikander Saifullah

Devices and techniques are generally described for providing a graphical user interface. In some examples, first image data may be sent to a computing device, that when rendered is effective to display a first combination of items disposed together in a first environment and a first selectable control. In some examples, first data may be received indicating a selection of the first selectable control. In various examples, a first computing device may determine first feature data representing the first image data. In some examples, the first computing device may determine second image data using the first feature data. In some examples, the second image data may be sent to a second computing device. The second image data, when rendered, may be effective to display a second combination of items disposed together in a second environment.

Visual search and content display system

Techniques are generally described for generation of photorealistic synthetic image data. A generator network generates first synthetic image data. A first class of image data represented by a first portion of the first synthetic image data is detected and the first portion is sent to a first discriminator network. The first discriminator network generates a prediction of whether the first portion of the first synthetic image data is synthetically generated. A second class of image data represented by a second portion of the first synthetic image data is detected and the second portion is sent to a second discriminator network. The second discriminator network generates a prediction of whether the second portion of the first synthetic image data is synthetically generated. The generator network is updated based on the predictions of the discriminators.

Generation of synthetic image data

Methods and systems for providing dynamic and category specific search suggestions are provided. According to one embodiment, a partial search query entered by a user can be used to determine a set of relevant search queries, which can be displayed as search suggestions within a search query suggestion window. Each query in the list can be associated with at least one specific category. According to various embodiments, the set of relevant queries and associated categories are dynamically updated as the user modifies the partial search query. Furthermore, each category is selectable by the user such that a detailed list of items associated only with the selected category can be displayed to the user.

Dynamic search suggestion and category specific completion

Disclosed are various embodiments for synchronizing interactive content with a live video stream. Items featured or discussed during a segment of a live video stream are determined. A manifest is generated for the segment. The manifest includes locator information for obtaining the segment and item information indicating the items. The manifest is then sent to one or more clients.

Synchronizing interactive content with a live video stream

Disclosed are various embodiments for navigating a taxonomy using search queries. An unstructured search query is parsed in order to identify one or more categories of a taxonomy of a collection of items. Each one of the one or more categories is associated with a confidence score. When the confidence score meets a threshold, a pool of items is selected from a respective one of the one or more categories. When no confidence score meets the threshold, the pool of items is selected from the collection of items.

Navigating a taxonomy using search queries

Disclosed are various embodiments for predictive page loading. Text entered by a user into a search query form is captured. If an evaluation of the size of the captured text or identity of characters in the captured text in accordance with a confidence criteria indicates that the confidence criteria is met, a speculative search query is provided to a server. The speculative search query includes a selected suggested search term. A received plurality of speculative search results is rendered in a hidden portion of a browser window. In response to receiving a user instruction to execute a committed search query that includes the at least one selected suggested search term, at least a portion of the received plurality of speculative search results is rendered in a visible portion of the browser window.

Predictive page loading based on text entry and search term suggestions

Disclosed are various embodiments for predictive page loading. A number of speculative search queries are provided to a server. Each speculative search query includes one of a plurality of suggested keywords. The number of speculative search queries is based on the number of suggested keywords. Responses corresponding to the speculative search queries are obtained. Each response includes a corresponding plurality of speculative search results. A portion of speculative search results from more than one of the responses is rendered in a hidden portion of a browser window. The portion of the speculative search results is rendered in a visible portion of the browser window in response to receiving a user instruction to execute a committed search query that includes a suggested keyword in one of the speculative queries.

Predictive page loading based on suggestion data

Disclosed are various embodiments for predictive network page loading. Content corresponding to a network request is obtained. A next network page associated with the obtained page content is predicted. If the prediction is confident relative to a predetermined confidence threshold, then a first network page is generated. The first network page includes the page content corresponding to the network request and a portion of content for the predicted network page. The generated first network page is provided to a client. A request for another network page is received. If the another network page in the request corresponds to the predicted next network page, a second network page is generated. The second network page includes the remainder of the content for the predicted page content.

Predictive page loading based on navigation

Effective color-aware search of a collection of content associated with one or more images is enabled. Content and/or its associated images may be automatically associated with representative palette colors in a suite of color palettes. Color palettes may be of a variety of types and have a hierarchical structure in which lower levels enable increasingly subtle distinctions between shades of color. Color palette hierarchies may be effectively presented, and appropriate portions emphasized based on associated search result sets. Search result sets may be refined and/or reordered in accordance with color palette selections and/or representative confidences of color palette selections for items at least referenced therein.

Color palette maps for color-aware search

Interactive color palettes for color-aware search

Growing community of inventors

Seattle, WA, United States of America

Alexandru I Indrei