公开(公告)号：US20220172065A1

公开(公告)日：2022-06-02

申请号：US17531985

申请日：2021-11-22

申请人： Mercari, Inc.

发明人： Sho ARORA ,  Jeffrey Kenichiro HARA ,  Sahil RISHI ,  Yu ISHIKAWA ,  Shotaro KOHAMA ,  Lu SUN ,  Vishal KASHYAP ,  Mohammad-Mahdi MOAZZAMI

IPC分类号： G06N3/08 ,  G06Q10/08

摘要： Disclosed herein are system, computer-readable storage medium, and method embodiments of automatic ontology generation by embedding representations. A system including at least one processor may be configured to receive a vectorized feature set derived from an embedding and including first and second features, and provide the vectorized feature set to a fuser set including first and second fusers. The system may be configured to generate a representation from the fuser set based on the first and second features, and derive tasks based on the representation, assigning to the tasks respective qualifier sets including a weight value, a loss function, and a feedforward function. The system may be configured to compute respective weighted losses for the tasks, based on the respective qualifier sets, and output a data model based on backpropagating the respective weighted losses through the fuser set, the vectorized feature set, the embedding, or a combination thereof.

公开(公告)号：US20210357955A1

公开(公告)日：2021-11-18

申请号：US17302581

申请日：2021-05-06

申请人： Mercari, Inc.

发明人： Sahil RISHI ,  Manikandan SANKAR ,  Byong Mok OH ,  Yodhavee CHUENBUNLUESOOK ,  Shuichi IIDA ,  Jeffrey Kenichiro HARA ,  Stephen JOHNSON

IPC分类号： G06Q30/02 ,  G06F16/957 ,  G06F17/18

摘要： Described herein are embodiments for improving search engine results of listings of For Sale Objects (FSOs). A search engine may be improved by implementing rules that resolve ambiguity between listings for different (FSOs) that match the same search inputs. An unsupervised machine learning module may evaluate candidate rules and identify improvements that may not be obvious to a human evaluator. An ecommerce site that combines the improved search engine with the unsupervised machine learning module may dynamically evaluate search results using different candidate rules and iteratively improve search results.