Abstract: A speech processing engine is provided that in some embodiments, employs Kalman filtering with a particular speaker's glottal information to clean up an audio speech signal for more efficient automatic speech recognition.

Abstract: Usages of language are analyzed in ways that are at least partially language independent. In preferred embodiments, portions of a document are hashed, and the resulting hash values are compared with each other and with those of other documents in real-time. Analyses can be used to gauge conformity of a document to one or more standards utilizing a hash triplet consisting of a hash for each document word and two involving the word and its preceding and following words, to provide suggestions to the author, and to filter email.

Abstract: Methods for converting a natural language sentence into a set of primitive sentences. The method includes identifying verbal blocks in the sentence, splitting the sentence into a set of logical clauses, determining a type of each clause wherein the type of each clause determines whether or not said clause contains verbal blocks, disambiguating ambiguous verbal blocks within each logical Clause, mad constructing a primitive sentence for each verbal block by duplicating the shared noun phrases of verbal blocks.

Abstract: A sequence is received of time domain digital audio samples representing sound (e.g., a sound generated by a human voice or a musical instrument). The time domain digital audio samples are processed to derive a corresponding sequence of audio pulses in the time domain. Each of the audio pulses is associated with a characteristic frequency. Frequency domain information is derived about each of at least some of the audio pulses. The sound represented by the time domain digital audio samples is transformed by processing the audio pulses using the frequency domain information. The sound transformation utilizes overlapping windows and a computational cost function which depends on a product of the number of the pitch periods and the inverse of the minimum fundamental frequency within the window is determined.

Abstract: A multi-language sorting index may allow sort keys for multiple languages to be stored in an auxiliary table associated with a primary table in which the strings to be sorted are stored. Sort keys may be generated for each string and stored in the auxiliary table, with duplicate sort keys excluded. Queries for sorted data in the first table may use the auxiliary table to identify and sort the requested data.

Abstract: An apparatus includes a translation memory storing previously translated sentences in association with source sentences; a matching rate calculation unit comparing original sentences in an original text to be translated with the source sentences in the translation memory and thereby calculating matching rates of the original sentences to the source sentences; a counting unit counting numbers of words for the respective matching rates; an adjusting unit adjusting the numbers of words of the respective matching rates by multiplying the numbers of words by the corresponding coefficients defined in a coefficient table; a total-no-of-words calculation unit adding the adjusted numbers of words to obtain a total number of words of the original text; and a delivery period setting unit obtaining a value corresponding to the obtained total number of words from a delivery period table and setting the value as a delivery period for a translation of the original text.

Abstract: The disclosed embodiments relate to a system that corrects user-interface shift errors. During operation of the system, a user provides an input through a user interface. In response, the system compares the input to a pre-determined set of phrases associated with user-interface errors. In particular, the user-interface errors may include shift errors in a position of a finger or a hand of the user relative to the user interface. Based on the comparisons, the system identifies a correct input, thereby facilitating correction of the shift error that occurred when the user provided the input through the user interface. Next, the system provides feedback to the user. For example, the feedback may correct the received input or may provide a suggestion to the user that includes the suggested correct input.

Abstract: A foreign language writing service method includes: recognizing, when a mixed text of foreign language portions and mother tongue portions is entered by a learner, the mother tongue portions from the mixed text; translating the mother tongue portions; combining a mother tongue translation result with the foreign language portions of the mixed text to generate a combined text; and providing the learner with the combined text of the mother tongue translation result and the foreign language portions of the mixed text.

Abstract: Methods, systems, and media are provided for identifying languages in multilingual text. A document is decoded into a universal representative coding for easier tag manipulation, then broken into plain-text content sections. The sections are identified and assigned a weight, wherein more informative sections are given a higher weight and less informative sections are given a lesser weight. A language likelihood score is determined for each word, phrase, or character n-gram in a section. The language likelihood scores within a section are combined for each language. The combined section scores are then summed together to obtain a total document score for each language. This results in a document score for each language, which can be ranked to determine the primary language for the document.

Abstract: A method for producing healthcare data records from graphical inputs by computer users. Includes generating a plurality of user input categories, displaying on a graphical display icons that correspond to a first of the user input categories and receiving a first user selection of a first icon of the plurality of icons, and displaying on the graphical display a plurality of icons that correspond to a second of the user input categories and receiving a second user selection of a second icon of the plurality of icons. The method also includes displaying icons that correspond to a physical target on which the medical action or observation is performed and receiving a third user selection of the physical target, and applying a syntax to populate a data record of the action using the at least two of the first, second, and third user selections.

Abstract: A frequency-domain noise shaping method and device interpolates a spectral shape and a time-domain envelope of a quantization noise in a windowed and transform-coded audio signal. In the method and device, transform coefficients of the windowed and transform-coded audio signal are split into a plurality of spectral bands. For each spectral band, a first gain representing a spectral shape of the quantization noise at a first transition between a first time window and a second time window is calculated, a second gain representing a spectral shape of the quantization noise at a second transition between the second time window and a third time window is calculated, and the transform coefficients of the second time window are filtered based on the first and second gains, to interpolate between the first and second transitions the spectral shape and the time-domain envelope of the quantization noise.

Abstract: Systems, methods, apparatuses and computer program products configured to provide sound feedback for input devices are described. Embodiments take input from a digitizer, such as input using as stylus/pen, and produce sound feedback to enhance the user's input interface experience. Embodiments thus provide a user with a more realistic interface with an electronic device, emulating use of conventional writing implements.

Abstract: Disclosed herein are systems, methods, and non-transitory computer-readable storage media for generating a model for use with automatic speech recognition. These principles can be implemented as part of a streamlined tool for automatic training and tuning of speech, or other, models with a fast turnaround and with limited human involvement. A system configured to practice the method receives, as part of a request to generate a model, input data and a seed model. The system receives a cost function indicating accuracy and at least one of speed and memory usage, The system processes the input data based on seed model and based on parameters that optimize the cost function to yield an updated model, and outputs the updated model.

Abstract: A disclosed speech sound enhancement device includes an SNR calculation unit for calculating an SNR which is a ratio of received speech sound to ambient noise; a first-frequency-range enhancement magnitude calculation unit for calculating, based on the SNR and frequency-range division information indicating a first and a second frequency range, enhancement magnitude of the first frequency range, the first frequency range contributing to an improvement of subjective intelligibility of the received speech sound, the second frequency range contributing to an improvement of subjective articulation of the received speech sound; a second-frequency-range enhancement magnitude calculation unit for calculating enhancement magnitude of the second frequency range based on the enhancement magnitude of the first frequency range; and a spectrum processing unit for processing spectra of the received speech sound using the enhancement magnitude of the first frequency range, the enhancement magnitude of the second frequency r

Abstract: Methods and a system for a natural language control interface are provided to enable a user to modify various properties of a document. The modifications comprise building sentences from modification words, and combining them together in one display. The modifications are displayed in real time for a user to observe as they are inputted. The order of the modifications is managed by the user and is configured to be changed, added and/or removed.

Abstract: Systems and methods are operable to associate each of a plurality of stored audio patterns with at least one of a plurality of digital tokens, identify a user based on user identification input, access a plurality of stored audio patterns associated with a user based on the user identification input, receive from a user at least one audio input from a custom language made up of custom language elements wherein the elements include at least one monosyllabic representation of a number, letter or word, select one of the plurality of stored audio patterns associated with the identified user, in the case that the audio input received from the identified user corresponds with one of the plurality of stored audio patterns, determine the digital token associated with the selected one of the plurality of stored audio patterns, and generate the output signal for use in a device based on the determined digital token.

Abstract: A method of performing a search of a database of speakers, includes: receiving a query speech sample spoken by a query speaker; deriving a query utterance from the query speech sample; extracting query utterance statistics from the query utterance; performing Kernelized Locality-Sensitive Hashing (KLSH) using a kernel function, the KLSH using as input the query utterance statistics and utterance statistics extracted from a plurality of utterances included in a database of speakers in order to select a subset of the plurality of utterances; and comparing, using an utterance comparison equation, the query utterance statistics to the utterance statistics for each utterance in the subset to generate a list of speakers from the database of utterances having a highest similarity to the query speaker.

Abstract: A method, computer program product, and system are provided for correcting one or more typed words on an electronic device. The method can include receiving one or more typed words from a text input device and generating one or more candidate words for the one or more typed words. The method can also include receiving an audio stream at the electronic device that corresponds to the one or more typed words. The audio stream can then be translated into text using the one or more candidate words, where the translation includes assigning a confidence score to each of the one or more candidate words. Based on the confidence score associated with each of the one or more candidate words, a candidate word can be selected among the one or more candidate words to represent each portion of the text. A word from the one or more typed words can be replaced with the selected candidate word based on the value of the confidence score associated with the selected candidate word.

Abstract: Variation over time in fundamental frequency in singing voices is separated into a melody-dependent component and a phoneme-dependent component, modeled for each of the components and stored into a singing synthesizing database. In execution of singing synthesis, a pitch curve indicative of variation over time in fundamental frequency of the melody is synthesized in accordance with an arrangement of notes represented by a singing synthesizing score and the melody-dependent component, and the pitch curve is corrected, for each of pitch curve sections corresponding to phonemes constituting lyrics, using a phoneme-dependent component model corresponding to the phoneme. Such arrangements can accurately model a singing expression, unique to a singing person and appearing in a melody singing style of the person, while taking into account phoneme-dependent pitch variation, and thereby permits synthesis of singing voices that sound more natural.

Abstract: Automated question answering is disclosed that relates to the selection of an answer to a question from a pool of potential answers which are manually or automatically extracted from a large collection of textual documents. The a feature extraction component, a feature combination component, an answer selection component, and an answer presentation component, among others, are included. The input to the system is a set of one or more natural language questions and a collection of textual document. The output is a (possibly ranked) set of factual answers to the questions, these answers being extracted from the document collection.