Google Inc.

Macduff Richard Hughes

Yonghui Wu

Zhifeng Chen

Niki J Parmar

Wolfgang Macherey

Melvin Jose Johnson Premkumar

Michael Schuster

Llion Owen Jones

Ankur Bapna

Orhan Firat

Xu Chen

George Foster

Alexander Jay Cuthbert

Wei Wang

Sunny Goyal

Joshua James Estelle

Taro Watanabe

Bowen Liang

Tetsuji Nakagawa

Alexander Rudnick

Minqi Sebastian Jiang

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for machine translation using neural networks. In some implementations, a text in one language is translated into a second language using a neural network model. The model can include an encoder neural network comprising a plurality of bidirectional recurrent neural network layers. The encoding vectors are processed using a multi-headed attention module configured to generate multiple attention context vectors for each encoding vector. A decoder neural network generates a sequence of decoder output vectors using the attention context vectors. The decoder output vectors can represent distributions over various language elements of the second language, allowing a translation of the text into the second language to be determined based on the sequence of decoder output vectors.

Machine translation using neural network models

A method includes generating a base model by training with a first dataset of data pairs and generating an adapted model by training the base model on a second dataset of data pairs. The method also includes determining a contrastive score for each data pair of a third dataset of data pairs using the base model and the adapted model. The contrastive score is indicative of a probability of quality of the respective data pair. The method also includes training a target model using the data pairs of the third dataset and the contrastive scores.

Contrastive sequence-to-sequence data selector

A computer-implemented technique includes techniques are presented for user image capture feedback for improved machine language translation. When machine language translation of OCR text obtained from an initial image has a low degree of likelihood of being an appropriate translation, these techniques provide for user image capture feedback to obtain additional images to obtain a modified OCR text, which can result in improved machine language translation results. Instead of user image capture feedback, the techniques may obtain the modified OCR text by selecting another possible OCR text from the initial OCR operation. In addition to additional image capturing, light source intensity and/or a quantity/number of light source flashes can be adjusted. After obtaining the modified OCR text, another machine language translation can be obtained and, if it has a high enough degree of likelihood, it can then be output to a user.

Techniques for providing user image capture feedback for improved machine language translation

A data processing apparatus receives data indicating a movement of a client device by a first user. The apparatus determines that the movement of the client device is a delimiter motion for switching between a first mode, in which the client device is configured to (i) provide a first interface for a first user speaking in a first language and (ii) perform speech recognition of the first language, and a second mode, in which the client device is configured to (i) provide a second interface for a second user speaking in a second language and (ii) perform speech recognition of the second language, the second interface being different from the first interface. Based on determining that the movement is a delimiter motion, the apparatus switches between the first mode and the second mode without the second user physically interacting with the client device.

Motion responsive user interface for realtime language translation

Growing community of inventors

Los Gatos, CA, United States of America

Macduff Richard Hughes