Google Inc.

Matthew I Lloyd

Trausti T Kristjansson

Pankaj Risbood

Johan Schalkwyk

Willard Van Tuyl Rusch, Ii

Martin Jansche

Stanley Chen

Nicholas Lynn

Gregory Joseph Badros

Douglas H Beeferman

Doug Beeferman

Michael Hochberg

Thomas Bergan

Willard Van Tuyl Rusch Ii

Implementations include a method, systems, and computer program products for adaptively configuring codecs to transmit encoded data to better utilize available bandwidth of reachable networks. One implementation is a method of transmitting encoded data from a host to a target. The method includes, selecting a communication network reachable by the host, estimating an available bandwidth from the host to the target via the selected communication network, configuring a codec from a plurality of codecs implemented in the host according to the estimated available bandwidth, encoding data using the configured codec, and transmitting the encoded data to the target.

Matching encoder output to network bandwidth

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for processing spoken query terms. In one aspect, a method includes performing speech recognition on an audio signal to select two or more textual, candidate transcriptions that match a spoken query term, and to establish a speech recognition confidence value for each candidate transcription, obtaining a search history for a user who spoke the spoken query term, where the search history references one or more past search queries that have been submitted by the user, generating one or more n-grams from each candidate transcription, where each n-gram is a subsequence of n phonemes, syllables, letters, characters, words or terms from a respective candidate transcription, and determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries, and a weighting value that is based on the respective frequency.

Disambiguation of a spoken query term

A method may include storing, in a memory associated with a computer device, cached content associated with a document and receiving a request to access the document. The method may further include sending a request to revalidate the cached content to a server device associated with the document, performing speculative rendering of the document, using the cached content, the speculative rendering being performed as a plurality of work items, receiving revalidated content associated with the document, determining invalid ones of the plurality of work items based on the revalidated content, and discarding the invalid ones of the plurality of work items.

Speculative rendering during cache revalidation

Methods, systems, and apparatuses, including computer programs encoded on a computer storage medium, for adjusting language models. In one aspect, a method includes accessing audio data. Information that indicates a first context is accessed, the first context being associated with the audio data. At least one term is accessed. Information that indicates a second context is accessed, the second context being associated with the term. A similarity score is determined that indicates a degree of similarity between the second context and the first context. A language model is adjusted based on the accessed term and the determined similarity score to generate an adjusted language model. Speech recognition is performed on the audio data using the adjusted language model to select one or more candidate transcriptions for a portion of the audio data.

Adjusting language models based on topics identified using context

A computer-implemented method includes receiving a plurality of uniform resource locators (URLs), where the URLs identify content files, and where the URLs include at least one parameter. Fingerprints of the content files are identified. A first entropy of values of the fingerprints conditional on values of a first parameter is determined, and a second entropy of values of the first parameter conditional on values of the fingerprints is determined. It is determined that the first parameter is irrelevant to the identification of a unique content file by the URLs based, at least in part, on the first and second entropy values.

Uniform resource locator canonicalization

Adjusting language models using context information

The present disclosure includes processing a signal to generate a first sub-set of data, transmitting the first sub-set of data for generation of a reconstructed audio signal, the reconstructed audio signal having a fidelity relative to the signal, processing the signal to generate a second sub-set of data and a third sub-set of data, the second sub-set of data defining a second portion of the signal and comprising data that is different than data of the first sub-set of data, and the third sub-set of data defining a third portion of the signal and comprising data that is different than data of the first and second sub-sets of data, comparing a priority of the second sub-set of data to a priority of the third sub-set of data, and transmitting one of the second sub-set of data and the third sub-set of data over the network for improving the fidelity.

Progressive encoding of audio

The disclosure generally relates to identifying gestures on mobile devices for use in connection with speech detection. A gesture may be identified using at least one Bayesian network, such as a Hidden Markov Model. Each Bayesian network corresponds to one of a plurality of different gesture types, which may include different gestures or similar gestures performed at different speeds, for example. Each Bayesian network includes states corresponding to partitioned segments in an accelerometer time series for the associated gesture type. Segmenting the accelerometer timeseries allows the Bayesian network to account for the speed at which the user makes the gesture, minimizing any effect of the user's speed in performing the gesture on identifying the gesture type. An action may be taken based on the identified gesture type. For example, speech recording may start or stop and/or a notification regarding speech recording may be provided depending on the gesture identified.

Behavioral recognition on mobile devices

A device may generate, based on historical access pattern information of web content items, a set of web content items that are likely to be accessed by a first user of a client device through a browser program. The device may transmit the determined set of web content items to the client device for pre-population of the cache of the browser program.

Browser cache pre-population

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for disambiguating contact information. A method includes receiving an audio signal, generating an affinity score based on a frequency with which a user has previously communicated with a contact associated with an item of contact information, and further based on a recency of one or more past interactions between the user and the contact associated with the item of contact information, inferring a probability that the user intends to initiate a communication using the item of contact information based on the affinity score generated for the item of contact information, and generating a communication initiation grammar.

Disambiguation of contact information using historical data

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for disambiguating contact information are described. A method includes determining, for each of multiple communications that were initiated by a user of a mobile device, a time when the communication was initiated or received; determining, for each of multiple contacts associated with the user, a probability associated with the contact based at least on the times when the communications were initiated or received; weighting a contact disambiguation grammar according to the probabilities; and processing audio data using the contact disambiguation grammar to select a particular contact.

Disambiguation of contact information using historical and context data

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for enhancing speech recognition accuracy. In one aspect, a method includes receiving an audio signal that corresponds to an utterance recorded by a mobile device, determining a geographic location associated with the mobile device, identifying a set of geotagged audio signals that correspond to environmental audio associated with the geographic location, weighting each geotagged audio signal of the set of geotagged audio signals based on metadata associated with the respective geotagged audio signal, and using the set of weighted geotagged audio signals to perform noise compensation on the audio signal that corresponds to the utterance.

Geotagged environmental audio for enhanced speech recognition accuracy

An audio signal generated by a device based on audio input from a user may be received. The audio signal may include at least a user audio portion that corresponds to one or more user utterances recorded by the device. A user speech model associated with the user may be accessed and a determination may be made background audio in the audio signal is below a defined threshold. In response to determining that the background audio in the audio signal is below the defined threshold, the accessed user speech model may be adapted based on the audio signal to generate an adapted user speech model that models speech characteristics of the user. Noise compensation may be performed on the received audio signal using the adapted user speech model to generate a filtered audio signal with reduced background audio compared to the received audio signal.

Growing community of inventors

Cambridge, MA, United States of America

Matthew I Lloyd