Google Inc.

Dmitriy Genzel

Franz Josef Och

Ashok Chhabedia Popat

Thomas Deselaers

Daniel M Keysers

Michael Jahr

Ignacio E Thayer

Ioannis Tsochandaridis

Methods and systems for grapheme splitting of text input for recognition are provided. A method may include receiving a text input in a script and segmenting the text input into one or more graphemes. Each of the one or more graphemes may be split into one or more recognition units based on one or more recognition unit identification criteria associated with the script. Next, a text recognition system may be trained using the recognition units. Text input may be handwritten text input received from a user or a scanned image of text.

Text recognition based on recognition units

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

Methods, systems, and apparatus, including computer program products, for correcting spelling in text. A text input is received for translation. One or more suspect words in the text input are identified. For each suspect word, one or more candidate words are identified. A score for the text input and scores for each of one or more candidate inputs are determined, where each candidate input is the text input with one or more of the suspect words each replaced by a respective candidate word. If any, a candidate input whose score is highest among the scores for the candidate inputs and is greater than the text input score by at least a threshold is selected. Otherwise, the text input is selected. A translation of a selected candidate input or the selected text input is provided as the translation of the text input.

Automatic spelling correction for machine translation

Systems, methods, and apparatuses including computer program products are provided for training machine learning systems. In some implementations, a method is provided. The method includes receiving a collection of phrases, normalizing a plurality of phrases of the collection of phrases, the normalizing being based at least in part on lexicographic normalizing rules, and generating a normalized phrase table including a plurality of key-value pairs, each key value pair includes a key corresponding to a normalized phrase and a value corresponding to one or more un-normalized phrases associated with the normalized key, each un-normalized phrase having one or more parameters.

Growing community of inventors

Mountain View, CA, United States of America

Dmitriy Genzel