公开(公告)号：US08832091B1

公开(公告)日：2014-09-09

申请号：US13647296

申请日：2012-10-08

Applicant: Amazon Technologies, Inc.

Inventor： Rahul H. Bhagat ,  Brian Cameros ,  Srikanth Thirumalai

IPC: G06F7/00 ,  G06F17/30

CPC classification number: G06F17/30958

Abstract: The use of graph-based semantic analysis with respect to items and tags may enable the discovery of the characteristics of items. A tag collection component may initially obtain a corresponding set of tags for each item of multiple items. A graph generation component may then generate a graph that includes a corresponding item node for each item and a corresponding tag node for each tag cluster of tags. Each item node in the graph may be connected to each of one or more tag nodes by a respective edge. Subsequently, following assignment of a label to each tag node, a graph evaluation component may execute a random walk algorithm on the graph. The execution of the random walk algorithm may provide a corresponding ranked list of tags for each item or a corresponding set of correlated tags for each tag.

Abstract translation: 使用关于项目和标签的基于图形的语义分析可以使得能够发现项目的特征。 标签收集组件可以最初为多个项目的每个项目获得相应的一组标签。 然后，图形生成组件可以生成包括每个项目的相应项目节点以及每个标签集群的对应标签节点的图形。 图中的每个项目节点可以通过相应的边缘连接到一个或多个标签节点中的每一个节点。 随后，在将标签分配给每个标签节点之后，图形评估组件可以在图上执行随机游走算法。 随机游走算法的执行可以为每个项目或每个标签的相应标签的相应集合提供相应的排名列表的标签。

公开(公告)号：US09615136B1

公开(公告)日：2017-04-04

申请号：US13887107

申请日：2013-05-03

Applicant: Amazon Technologies, Inc.

Inventor： Grant Michael Emery ,  Rahul Hemant Bhagat ,  Brian Cameros ,  Benjamin Thomas Cohen ,  Logan Luyet Dillard ,  Yongwen Liang ,  Scott Allen Mongrain ,  Michael David Quinn ,  Eli Glen Rosofsky ,  Adam Callahan Sanders

IPC: H04N7/173 ,  H04N21/472 ,  H04N21/258 ,  H04N21/442 ,  H04N5/445

CPC classification number: H04N21/47202 ,  H04N21/25891 ,  H04N21/44222 ,  H04N21/4756 ,  H04N21/4828

Abstract: The present technology may identify item category affinities by identifying a plurality of classifications of an item. An accuracy of the plurality of classifications relative to one another for the item may be identified. A category affinity of the item may be determined based on the accuracy of the plurality of classifications relative to one another.