Amazon Technologies, Inc.

Anirban Sengupta

Shaunak Mishra

Ravendar Lal

Xidong Pi

Phani Teja Anumanchupallik

Chaoran Wei

Sarat Chandra Vadapalli

Techniques for providing information for presentation at a device are described. In an example, a computer system receives, from a device, a query indicating that item information is requested. The computer system determines a topic for the item information. The computer system determines a sub-topic of the topic, the sub-topic generated based at least in part on a semantic overlap between predefined item attribute refinements and user contributed entries related to the item. The computer system causes the device to present the item information, the topic, and the sub-topic in association with the topic.

Topics and sub-topics for presenting search results on a device

Systems and methods for trending product recommendations are provided herein. Particularly, click-through-rate (CTR) and conversion-rate (CVR) data metrics associated with various products may be analyzed to identify trending products. Keyphrases may then be extracted from information associated with these trending products and provided to an unsupervised graph-based model to determine a relative importance of each keyphrase relative to other keyphrases extracted from information associated with other trending products. Keyphrases that are determined to be the most 'important' may then be presented along with the trending products in a trending product widget when a user searches for products.

Systems and methods for trending product recommendations

Disclosed are various embodiments for inferring brand similarities using graph neural networks and selection prediction. In one embodiment, a brand-to-brand graph is generated indicating similarities between a set of brands according to at least one of: click-through data or conversion data. Using a first graph neural network (GNN) tower, the brand-to-brand graph is analyzed to determine brand similarities among a first brand identified from a search query and a first set of other brands. Using a second GNN tower, the brand-to-brand graph is analyzed to determine brand similarities among a second brand and a second set of other brands. A level of similarity between the first brand and the second brand is determined based at least in part on an output of the first GNN tower and an output of the second GNN tower.

Growing community of inventors

Sammamish, WA, United States of America

Anirban Sengupta