Abstract: A machine readable storage can include a set of instructions for load balancing. The storage can include a plug-in receptor of a load balancer. The plug-in receptor can be compliant with a known industry standard and can be is associated with a two or more load balancing algorithms. The load balancer can utilize selected ones of the load balancing algorithms to determine which of two or more voice servers are to handle incoming speech processing requests. Selected ones of the load balancing algorithms can include a speech utilization algorithm. The speech utilization algorithm can calculate a speech utilization score for at least one of the voice servers based upon speech processing specific factors.

Abstract: Methods, devices, and systems for coding and decoding audio are disclosed. At least two transforms are applied on an audio signal, each with different transform periods for better resolutions at both low and high frequencies. The transform coefficients are selected and combined such that the data rate remains similar as a single transform. The transform coefficients may be coded with a fast lattice vector quantizer. The quantizer has a high rate quantizer and a low rate quantizer. The high rate quantizer includes a scheme to truncate the lattice. The low rate quantizer includes a table based searching method. The low rate quantizer may also include a table based indexing scheme. The high rate quantizer may further include Huffman coding for the quantization indices of transform coefficients to improve the quantizing/coding efficiency.

Abstract: A novel bandwidth extension technique allows information to be encoded and decoded using a fractal self similarity model or an accurate spectral replacement model, or both. Also a multi-band temporal amplitude coding technique, useful as an enhancement to any coding/decoding technique, helps with accurate reconstruction of the temporal envelope and employs a utility filterbank. A perceptual coder using a comodulation masking release model, operating typically with more conventional perceptual coders, makes the perceptual model more accurate and hence increases the efficiency of the overall perceptual coder.

Abstract: An improved handheld electronic device and associated method employing an improved spell checking routine enable proposed spelling corrections having a close logical proximity to an active input to be output at a position of preference for easy selection by the user. By way of example, a base character and the various accented forms thereof can be said to have a logical proximity to one another that is closer than their logical proximity to any character having a different base character, whether additionally having a diacritical element or not.

Abstract: A system and method recognizes speech securely using a secure multi-party computation protocol. The system includes a client and a server. The client is configured to provide securely speech in a form of an observation sequence of symbols, and the server is configured to provide securely a multiple trained hidden Markov models (HMMs), each trained HMM including a multiple states, a state transition probability distribution and an initial state distribution, and each state including a subset of the observation symbols and an observation symbol probability distribution. The observation symbol probability distributions are modeled by mixtures of Gaussian distributions. Also included are means for determining securely, for each HMM, a likelihood the observation sequence is produced by the states of the HMM, and means for determining a particular symbol with a maximum likelihood of a particular subset of the symbols corresponding to the speech.

Abstract: An advanced model that includes new processes is provided for use as a component of an effective disfluency identifier. The disfluency identifier tags edited words in transcribed speech. A speech recognition unit in combination with a part-of-speech tagger, a disfluency identifier, and a parser form a natural language system that helps machines properly interpret spoken utterances.

Abstract: A method of analyzing dialog between a user and an interactive application having dialog turns is provided. The method includes accessing information indicative of a plurality of dialog turns between the application and at least one user and identifying instances where the application determined a response was received before an associated prompt had completed. The accessed information includes information related to operation of the application with a first grammar to recognize the response. The method includes identifying whether the response was received in a particular limited time period from when the associated prompt began. If the response was received in the limited time period, the method determines whether the response included one or more terms from the associated prompt by performing recognition on the response using a second grammar having more information related to grammar of a language than the first grammar.

Abstract: In one embodiment, the method includes receiving an audio data frame having at least one channel. The channel is subdivided into a plurality of blocks, and at least two of the blocks are capable of different lengths. The embodiment further includes obtaining information from the audio signal indicating the subdivision of the channel into the blocks, and decoding the channel based on the obtained information. In one embodiment, an optimum prediction order is obtained for each block in the channel, where a prediction order indicates a number of prediction coefficients. The optimum prediction order indicates a minimum one of a global prediction order and a local prediction order. The global prediction order is determined based on a maximum permitted prediction order, and the local prediction order is determined based on a length of the block.

Abstract: A method and apparatus is disclosed for generating and distributing multilingual documents. The multilingual documents are comprised of primary information consisting of human-readable text and secondary information consisting of machine-readable data such that a translation of the text is accomplished by converting the human-readable text into a second language through the use of the decoded machine-readable data. The machine-readable data is comprised of a code that describes a set of editing operations that can be applied to the human-readable text to convert it into at least a second language. In a preferred embodiment, the machine-readable data is embedded in the image using an unobtrusive code on the document such as Xerox DATAGLYPH codes.

Abstract: A waveform interpolation speech coding apparatus and method for reducing complexity thereof are disclosed. The waveform interpolation speech coding apparatus includes: a waveform interpolation encoding unit for receiving a speech signal, calculating parameters for a waveform interpolation from the received speech signal, and quantizing the calculating parameters; and a realignment parameter calculating unit for restoring a characteristic waveform (CW) using the quantized parameter, calculating a realignment parameter that maximizes a cross-correlation among consecutive CWs for the restored CW.

Abstract: Parameters for a feature extractor and acoustic model of a speech recognition module are trained. An objective function is utilized to determine values for the feature extractor parameters and the acoustic model parameters.

Abstract: In one embodiment, distortion in a received speech signal is estimated using at least one model trained based on subjective quality assessment data. A speech quality assessment for the received speech signal is then determined based on the estimated distortion.

Abstract: Provided is a method for converting a dimension of a vector. The vector dimension conversion method for vector quantization includes the steps of: extracting a specific parameter having a pitch period from an input speech signal and then generating a vector of a dimension that varies according to the pitch period; dividing an entire frequency domain of the generated vector of the variable dimension into at least two frequency domains; and converting the vector of the variable dimension into vectors of mutually different fixed dimensions according to the divided frequency domains. Thereby, not only an error due to the vector dimension conversion is suppressed but codebook memory required for the vector quantization is effectively reduced.

Abstract: Techniques are disclosed for facilitating the process of proofreading draft transcripts of spoken audio streams. In general, proofreading of a draft transcript is facilitated by playing back the corresponding spoken audio stream with an emphasis on those regions in the audio stream that are highly relevant or likely to have been transcribed incorrectly. Regions may be emphasized by, for example, playing them back more slowly than regions that are of low relevance and likely to have been transcribed correctly. Emphasizing those regions of the audio stream that are most important to transcribe correctly and those regions that are most likely to have been transcribed incorrectly increases the likelihood that the proofreader will accurately correct any errors in those regions, thereby improving the overall accuracy of the transcript.

Abstract: Methods, systems, and apparatus, including computer program products, operable to perform operations including receiving from a user a search query; and receiving an indication of a user preference to apply transliteration in simplifying the query terms of the search query. Alternatively, the operations include receiving from a user a search query of query terms; applying transliteration in simplifying the query terms; and using the simplified query terms to identify synonyms to use in augmenting the search query. Alternatively, the operations include receiving from a user a search query; identifying the user interface language as a small language or not a small language; simplifying each query term to a simplified form; and if the user interface language is a small language, for each original query term that has a simplified form different from the original term, using the original query term as-is and not providing any synonyms for the query term.

Abstract: A method and apparatus for normalizing a histogram utilizing a backward cumulative histogram which can cumulate a probability distribution function in an order from a greatest to smallest value so as to estimate a noise robust histogram. A method of normalizing a speech feature vector includes: extracting the speech feature vector from a speech signal; calculating a probability distribution function using the extracted speech feature vector; calculating a backward cumulative distribution function by cumulating the probability distribution function in an order from a largest to smallest value; and normalizing a histogram using the backward cumulative distribution function.

Abstract: A method of automatically capitalizing text utilizes a capitalization model. The capitalization model is trained from data that is taken from documents associated with a particular user. In particular, documents that are authored by the user such as e-mails, are used to train the model.

Abstract: The disclosed technology is a computer application that establishes communication between a conversation-finder module and a computer application environment. The conversation-finder module determines a conversational floor based on three or more floor determination inputs. The conversational floor associates at least two of the three or more floor determination inputs as being on the conversational floor. The conversation-finder module and the computer application environment can be responsive to each other and can adapt to each other. That is (either or both), the computer application environment can adapt to the conversational floor(s) determined by the conversation-finder module; and the conversation-finder module can determine the conversational floors responsive to a floor determination input and/or control input from the computer application environment.

Abstract: A method of measuring the degree of enhancement made to a voice signal by receiving the voice signal, identifying formant regions in the voice signal, computing stationarity for each identified formant region, enhancing the voice signal, identifying formant regions in the enhanced voice signal that correspond to those identified in the received voice signal, computing stationarity for each formant region identified in the enhanced voice signal, comparing corresponding stationarity results for the received and enhanced voice signals, and calculating at least one user-definable statistic of the comparison results as the degree of enhancement made to the received voice signal.

Abstract: The present invention provides a method of processing at least one natural language text using a graph. The method includes determining a plurality of text units based upon the natural language text, associating the plurality of text units with a plurality of graph nodes, and determining at least one connecting relation between at least two of the plurality of text units. The method also includes associating the at least one connecting relation with at least one graph edge connecting at least two of the plurality of graph nodes and determining a plurality of rankings associated with the plurality of graph nodes based upon the at least one graph edge. The method can also include a graphical visualization of at least one important text unit in a natural language text or collection of texts. Methods for word sense disambiguation, keyword extraction, and sentence extraction are also provided.