Google Inc.

University of Southern California

Franz Josef Och

Peng Xu

Wolfgang Macherey

Jeffrey Adgate Dean

Thorsten Brants

Shankar Kumar

Ignacio E Thayer

Henry Allan Rowley

Jay Michael Ponte

Alexander Mark Franz

Jeffrey Chin

Daniel Rosart

John S Hawkins

Sha-Mayn Teh

Karel Driesen

Anton Carver

Yang Li

Ashok Chhabedia Popat

David Robert Talbot

Dmitriy Genzel

Thomas Deselaers

Jakob David Uszkoreit

Daniel M Keysers

Joshua James Estelle

Michael Jahr

Ashish Venugopal

Jurij Ganitkevic

Christopher James Dyer

Awaneesh Verma

Sarveshwar Duddu

Eahab Ibrahim

Roy W Tromble

Ioannis Tsochandaridis

Andrew M Dai

Klaus Macherey

Jeffrey Dean

Jay Ponte

John Hawkins

Systems, methods, and apparatus for accessing distributed models in automated machine processing, including using large language models in machine translation, speech recognition and other applications.

Encoding and adaptive, scalable accessing of distributed models

A computer-implemented method includes: receiving, at a user device, user input corresponding to handwritten text to be recognized using a recognition engine; and receiving, at the user device, a representation of the handwritten text. The representation includes the handwritten text parsed into individual handwritten characters. The method further includes: displaying, on a display of the user device, the handwritten characters using a first indicator; receiving, at the user device, an identification of a text character recognized as one of the handwritten characters; displaying, on the display, the text character; and adjusting, at the user device, the one of the handwritten characters from being displayed using the first indicator to using a second indicator in response to the received identification. The first and second indicators are different.

Progress display of handwriting input

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for decompounding compound words are disclosed. In one aspect, a method includes obtaining a token that includes a sequence of characters, identifying two or more candidate sub-words that are constituents of the token, and one or more morphological operations that are required to transform the sub-words into the token, where at least one of the morphological operations involves a use of a non-dictionary word, and determining a cost associated with each sub-word and a cost associated with each morphological operation.

Compound splitting

A system with a nonstatistical translation component integrated with a statistical translation component engine. The same corpus may be used for training the statistical engine and also for determining when to use the statistical engine and when to use the translation component. This training may use probabilistic techniques. Both the statistical engine and the translation components may be capable of translating the same information, however the system determines which component to use based on the training. Retraining can be carried out to add additional components, or when after additional translator training.

Method and system for translating information with a higher probability of a correct translation

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for performing optical character recognition. In one aspect, a method includes receiving a text image I. A set of feature functions are evaluated for a log linear model to determine respective feature values for the text image I, wherein each feature function hmaps the text image I to a feature value, and wherein each feature function his associated with a respective feature weight &#x3bb;. A transcription {circumflex over (T)} is determined that minimizes a cost of the log linear model.

Optical character recognition

Techniques described herein may recognize handwritten characters that are written at least partially over the top of one another that are input to a computing device. The handwritten characters may be formed of one or more strokes. A user may write characters or parts of words over approximately the same area of graphical user interface (i.e., on top of each other) without having to wait for a timeout between character input and without having to select a button or provide another input indicating the character is complete before entering input for another character. Once a character is at least partially recognized, a graphical indication corresponding to the user input displayed on a screen may be altered. Such alterations may include fading or changing size or location of the graphical indication.

Character recognition for overlapping textual user input

Systems, methods, and apparatuses including computer program products for machine translation. A method is provided that includes generating a plurality of machine translation systems using a single machine translation engine, and generating a consensus translation from a plurality of candidate translations for a source sentence, where each candidate translation of the plurality of candidate translations is an output of a respective machine translation system of the plurality of machine translation systems.

Consensus translations from multiple machine translation systems

A computer-implemented method includes receiving a document and a request to translate the document to a different language, the document including at least one tag associated with a first portion of text within the document, receiving a manual translation of the document translated by a human translator but not including the at least one tag, generating a plurality of alignments between the document and the manual translation using a statistical alignment model, selecting one of the plurality of alignments based on a likelihood that the first portion of text in the document corresponds to an aligned second portion of text within the manual translation, mapping a location of the tag in the document to a corresponding location within the manual translation based on the selected alignment, and inserting the at least one tag into the manual translation at the corresponding location to obtain a modified manual translation of the document.

Techniques for translating documents including tags

Systems, methods, and computer program products for machine translation are provided. In some implementations a system is provided. The system includes a language model including a collection of n-grams from a corpus, each n-gram having a corresponding relative frequency in the corpus and an order n corresponding to a number of tokens in the n-gram, each n-gram corresponding to a backoff n-gram having an order of n&#x2212;1 and a collection of backoff scores, each backoff score associated with an n-gram, the backoff score determined as a function of a backoff factor and a relative frequency of a corresponding backoff n-gram in the corpus.

Large language models in machine translation

A way of detecting a watermark present in a structured result, such as a search result or a machine translation. The structured result is received and a hash is computed based upon at least part of the result. The resulting bit sequence is tested against a null hypothesis that the bit sequence was generated by a random variable with a binomial distribution with a parameter p=0.5. The result of this test is compared to a significance level, which indicates whether the structured result is watermarked.

Watermarking of structured results and watermark detection

A way of generating a watermark for a structured result, such as a search result or a machine translation. A hash function is used to generate a bit sequence for each of a plurality of structured results. A ranking score is generated for each resulting bit sequence. The ranking score can be based on the detectability of the bit sequence compared to a randomly-generated bit sequence and the quality of each of the structured results. A structured result is selected as the watermarked structured result based upon the ranking score.

Watermarking of structured results and watermark generation

Methods, systems, and apparatus, including computer program products, for language translation are disclosed. In one implementation, a method is provided. The method includes accessing a hypothesis space, where the hypothesis space represents a plurality of candidate translations; performing decoding on the hypothesis space to obtain a translation hypothesis that minimizes an expected error in classification calculated relative to an evidence space; and providing the obtained translation hypothesis for use by a user as a suggested translation in a target translation.

Optimizing parameters for machine translation

Growing community of inventors

Palo Alto, CA, United States of America

Franz Josef Och