Microsoft Technology Licensing, LLC

Masaki Itagaki

Xuedong D Huang

Xiaodong He

Nanshan Zeng

Dimitrios Basile Dimitriadis

Yashesh Gaur

Takako Aikawa

Kenichi Kumatani

Robert Peter Gmyr

The disclosure herein describes training a global model based on a plurality of data sets. The global model is applied to each data set of the plurality of data sets and a plurality of gradients is generated based on that application. At least one gradient quality metric is determined for each gradient of the plurality of gradients. Based on the determined gradient quality metrics of the plurality of gradients, a plurality of weight factors is calculated. The plurality of gradients is transformed into a plurality of weighted gradients based on the calculated plurality of weight factors and a global gradient is generated based on the plurality of weighted gradients. The global model is updated based on the global gradient, wherein the updated global model, when applied to a data set, performs a task based on the data set and provides model output based on performing the task.

Dynamic gradient aggregation for training neural networks

A method of determining the consistency of training data for a machine translation system is disclosed. The method includes receiving a signal indicative of a source language corpus and a target language corpus. A textual string is extracted from the source language corpus. The textual string is aligned with the target language corpus to identify a translation for the textual string from the target language corpus. A consistency index is calculated based on a relationship between the textual string from the source language corpus and the translation. An indication of the consistency index is stored on a tangible medium.

Validation of the consistency of automatic terminology translation

Growing community of inventors

Redmond, WA, United States of America

Masaki Itagaki