Abstract: Methods, systems, and apparatus, including stored computer programs, for generating modified phonetic representations of Indic words. An Indic sequence of Indic character combinations that each include an Indic character is received. An orthographic representation, e.g., an English-orthographic representation, of the Indic sequence is received. The representation includes a character combination that has a consonant followed by vowel A phonetic representation, e.g., an International Phonetic Alphabet (IPA) representation, of the Indic sequence is generated and, based on a comparing, a schwa is removed from a character combination in the phonetic representation that corresponds to an orthographic character combination that does not have a consonant-vowel sequence.

Abstract: A system improves speech detection or processing by identifying registration signals. The system encodes a limited frequency band by varying the amplitude of a pulse width modulated signal between predefined values. The signal is separated into frequency bins that identify amplitude and phase. The registration signal is measured by comparing a difference in average acoustic power in a plurality of adjacent bins over time.

Abstract: The present invention relates to speech coding in wireless and wireline communication systems. The present invention provides a method of saving bandwidth by a controlled dropping of speech frames at an encoder in a sending communication device. The dropping is controlled in a manner to minimize the effects on the speech quality after the decoding in the receiving communication device, by assuring that the state mismatch between the encoder and the decoder is removed or at least significantly reduced. This is achieved by letting the encoder run an ECU algorithm with a similar behavior as the one running in the decoder in the receiving communication device.

Abstract: Provided are a device and method for interactive machine translation. The device includes a machine translation engine having a morphological/syntactic analyzer for analyzing morphemes and sentences of an original text and generating original text analysis information, and a translation generator for generating a translation and translation generation information on the basis of the original text analysis information, and a user interface module for displaying sentence structures of the original text and the translation, and a relationship between the original text and the translation to a user on the basis of the original text analysis information and the translation generation information, and for receiving corrections to the original text or the translation from the user. The device and method provide a user interface whereby the user can effectively recognize and correct a mistranslated part and a cause of the mistranslation, and rapidly provides a re-translated result according to the correction.

Abstract: Various embodiments of the present invention provide systems and methods for providing a translation for a set of one or more terms or phrases related to a software application using decentralized contributions. In particular, various embodiments provide systems and methods by which multiple users of the application contribute translations for individual terms or phrases of the application instead of having one entity supply the translation of the software application. Specifically, the process of various embodiments: (1) provides a repository for storing translations of the individual terms or phrases in a target language (e.g., a language for which a translation is needed); (2) collects the translations provided by users and stores the translations in the repository; (3) identifies the preferred translation for each individual term or phrase; and (4) displays the preferred translation to the user in response to receiving a user's request for a preferred translation in the target language.

Abstract: A noise reduction device comprises a SN ratio obtaining unit configured to obtain a SN ratio as a function of an estimated noise spectrum and an arithmetic product of an averaged power spectrum of the input signal and noise likeliness signal, and an output signal obtaining unit configured to obtain a output signal whose noise is reduced based on the input signal and the SN ratio obtained by the SN ratio obtaining unit.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for recognizing speech input. In one aspect, a method includes receiving a user input and a grammar including annotations, the user input comprising audio data and the annotations providing syntax and semantics to the grammar, retrieving third-party statistical speech recognition information, the statistical speech recognition information being transmitted over a network, generating a statistical language model (SLM) based on the grammar and the statistical speech recognition information, the SLM preserving semantics of the grammar, processing the user input using the SLM to generate one or more results, comparing the one or more results to candidates provided in the grammar, identifying a particular candidate of the grammar based on the comparing, and providing the particular candidate for input to an application executed on a computing device.

Abstract: Systems and methods are provided to automatically determine culture-based behavioral tendencies and preferences of individuals in the context of customer service interactions. For example, systems and methods are provided to process natural language dialog input of an individual to detect linguistic features indicative of individualistic and collectivistic behavioral tendencies and predict whether such individual will be cooperative or uncooperative with automated customer service.

Abstract: According to one embodiment, an electronic apparatus includes a sound characteristic output module configured to analyze audio data in video content data, thereby outputting sound characteristic information indicative of sound characteristics of the audio data. A talk section detection process module detects talk sections in which talks are made by persons, which are included in the video content data, on the basis of the sound characteristic information, and classifies the detected talk sections into a plurality of groups which are associated with different speakers. A display process module displays, on a time bar which is representative of a sequence of the video content data, a plurality of bar areas indicative of positions of the detected talk sections in the sequence of the video content data, in different display modes in association with the groups.

Abstract: Hierarchical coding of a source audio signal in the form of a data stream including a base level and at least two hierarchical enhancement levels, each of the levels being organized in successive frames. At least one frame of at least one enhancement level has a duration less than the duration of at least one frame of the base level. At least one indication representative of an order used for a set of enhancement level frames corresponding to the duration of at least one frame of the base level is inserted into the data stream.

Abstract: An encoding method and apparatus and a decoding method and apparatus are provided. The decoding method includes extracting a compatible down-mix signal optimized for a first multi-channel decoder from the input bitstream, converting the compatible down-mix signal to be optimized for a second multi-channel signal by performing a compatibility processing operation on the compatible down-mix signal, and generating a three-dimensional (3D) down-mix signal by performing a 3D rendering operation on the converted down-mix signal. Accordingly, it is possible to efficiently encode multi-channel signals with 3D effects and to adaptively restore and reproduce audio signals with optimum sound quality according to the characteristics of an audio reproduction environment.

Abstract: An encoding method and apparatus and a decoding method and apparatus are provided. The decoding method includes extracting a three-dimensional (3D) down-mix signal and spatial information from an input bitstream, removing 3D effects from the 3D down-mix signal by performing a 3D rendering operation on the 3D down-mix signal, and generating a multi-channel signal using the spatial information and a down-mix signal obtained by the removal. Accordingly, it is possible to efficiently encode multi-channel signals with 3D effects and to adaptively restore and reproduce audio signals with optimum sound quality according to the characteristics of a reproduction environment.

Abstract: A system and computer-implemented method for providing alternative voice interpretations to a user, including receiving an indication of a request for a search from a user comprising a voice input of search criteria, determining an n-best list of a plurality of possible interpretations of the voice input in response to receiving the indication, each possible interpretation comprising a word or phrase corresponding to the voice input, providing a first possible interpretation of the plurality of possible interpretations for display to the user, the first possible interpretation being the most likely interpretation of the voice input, receiving an indication that the first possible interpretation does not correspond to the voice input search criteria and providing one or more alternative interpretations of the plurality of interpretations for display to the user within a suggestion display area in response to receiving the indication.

Abstract: Speech is synthesized for a given text by determining a sequence of phonetic components based on the text, determining a sequence of target phonetic elements associated phonetic components, determining a sequence of target event types associated with the phonetic components and determining a sequence of speech units from a plurality of stored speech unit candidates by use of a cost function. The cost function comprises a unit cost, a concatenation cost, and an event type cost for each speech unit in the sequence of speech units. The unit cost of a speech unit is determined with respect to the corresponding target phonetic element, while the concatenation cost of a speech unit is determined with respect to adjacent speech units and the event type cost of each speech unit is determined with respect to the corresponding target event type.

Abstract: A signal bandwidth extension apparatus includes a determination unit which determines whether or not a peak component of the input signal is lacked in the band to be extended, and a control unit which controls to extend the bandwidth when the determination unit determines that the peak component of the input signal is lacked in the band to be extended, and not to extend the bandwidth when the determination unit determines that the peak component is not lacked.

Abstract: A method, program storage device and mobile device provide speech disambiguation. Audio for speech recognition processing is transmitted by the mobile device. Results representing alternates identified to match the transmitted audio are received. The alternates are displayed in a disambiguation dialog screen for making corrections to the alternates. Corrections are made to the alternates using the disambiguation dialog screen until a correct result is displayed. The correct result is selected. Content associated with the selected correct result is received in parallel with the receiving of the results representing alternates identified to match the transmitted audio.

Abstract: An encoding/decoding method, an coder/decoder (codec) and a radio communication device utilize a variable offset coding technique. In accordance with the technique, the start of processing of a first frame is time offset in relation to the end of the processing of the frame that precedes the first frame, the time offset bringing about a time gap between the end of the preceding frame and the start of processing the first frame. A substitution signal is inserted in the time gap.

Abstract: A method for variable resolution and error control in spoken language understanding (SLU) allows arranging the categories of the SLU into a hierarchy of different levels of specificity. The pre-determined hierarchy is used to identify different types of errors such as high-cost errors and low-cost errors and trade, if necessary, high cost errors for low cost errors.

Abstract: A system and method for semantic extraction using a neural network architecture includes indexing each word in an input sentence into a dictionary and using these indices to map each word to a d-dimensional vector (the features of which are learned). Together with this, position information for a word of interest (the word to labeled) and a verb of interest (the verb that the semantic role is being predicted for) with respect to a given word are also used. These positions are integrated by employing a linear layer that is adapted to the input sentence. Several linear transformations and squashing functions are then applied to output class probabilities for semantic role labels. All the weights for the whole architecture are trained by backpropagation.

Abstract: A system improves speech detection or processing by identifying registration signals. The system encodes a limited frequency band by varying the amplitude of a pulse width modulated signal between predefined values. The signal is separated into frequency bins that identify amplitude and phase. The registration signal is measured by comparing a difference in average acoustic power in a plurality of adjacent bins over time.