Abstract: A system configured to process an incoming spoken utterance and to coordinate among multiple speechlet components to execute an action of the utterance, where the output of one speechlet may be used as the input to another speechlet to ultimately perform the action. The speechlets and intervening actions need not be expressly invoked by the utterance. Rather the system may determine how best to complete the action and may identify intermediate speechlets that may be provide input data to the speechlet that will ultimately perform the action. The speechlets may be configured to recognize a common universe of actions and/or entities rather than have each speechlet or subject matter domain have its own set of recognizable actions and entities.

Abstract: A system and method for improving T-matrix training for speaker recognition, comprising receiving an audio input, divisible into a plurality of audio frames including at least an audio sample of a human speaker; generating for each audio frame a feature vector; generating for a first plurality of feature vectors centered statistics of at least a zero order and a first order; generating a first i-vector, the first i-vector representing the human speaker; and generating an optimized T-matrix training sequence computation, based on at least the first i-vector.

Abstract: Provided is an apparatus that includes a voice recognition section that executes a voice recognition process on a user speech and a learning processing section that executes a process of updating a degree of confidence on the basis of an interaction made between a user and the information processing apparatus after the user speech. The degree of confidence is an evaluation value indicating the reliability of a voice recognition result of the user speech. The voice recognition section generates data on degrees of confidence in recognition of the user speech in which data plural user speech candidates based on the voice recognition result of the user speech are associated with the degrees of confidence which are evaluation values each indicating reliability of the corresponding user speech candidate.

Abstract: A computer-implemented method according to one embodiment includes receiving, utilizing a processor, textual data associated with a conversation between a first participant and a second participant; receiving, utilizing the processor, an objective of the first participant for the conversation between the first participant and the second participant, where the objective is separate from the conversation; determining, utilizing the processor, a dialog act to be entered by the first participant that meets the objective, utilizing a model, including scoring a plurality of proposed dialog acts based on an amount that each proposed dialog act will change a probability of the objective being achieved during the conversation, and determining the dialog act to be entered, based on the scoring; and returning, utilizing the processor, the dialog act to the first participant.

Abstract: There is provided mechanisms for frame loss concealment. A method is performed by a receiving entity. The method comprises adding, in association with constructing a substitution frame for a lost frame, a noise component to the substitution frame. The noise component has a frequency characteristic corresponding to a low-resolution spectral representation of a signal in a previously received frame.

Abstract: A system and method are presented for the correction of packet loss in audio in automatic speech recognition (ASR) systems. Packet loss correction, as presented herein, occurs at the recognition stage without modifying any of the acoustic models generated during training. The behavior of the ASR engine in the absence of packet loss is thus not altered. To accomplish this, the actual input signal may be rectified, the recognition scores may be normalized to account for signal errors, and a best-estimate method using information from previous frames and acoustic models may be used to replace the noisy signal.

Abstract: Methods, apparatus and articles of manufacture for research data gathering are disclosed. An example apparatus disclosed herein is to detect whether the apparatus is powered by an internal power source or an external power source. The example apparatus is also to, in response to detecting the apparatus is powered by the internal power source, perform first processing on a received audio signal to determine audio data to store in storage of the apparatus. The example apparatus is further to, in response to detecting the apparatus is powered by the external power source, perform second processing on the stored audio data to recover the code, the second processing different from the first processing.

Abstract: Implementations of the disclosure are directed to OCR error correction systems and methods. In some implementations, a method comprises: obtaining, at a computing device, optical character recognition (OCR) text extracted from a document image, the text comprising a token; searching, at the computing device, based on a token bigram determined from the token and a mapping between words in a corpus and a corpus bigram set comprised of unique bigrams from the beginning or ending of the words in the corpus, the corpus for a best word to replace the token; and replacing, at the computing device, the token with the best word.

Abstract: Machine learning model training corpus debiasing includes identifying an attribute of input text selected from the training corpus, the attribute including word(s) of the input text, and the attribute corresponding to an attribute class encompassing different possible class values, recognizing bias in the input text with respect to the attribute class, and generating output text corresponding to the attribute and imparting diversity with respect to the attribute class and relative to the input text, where generating the output text uses an optimization function based on loss objectives to minimize loss in the generated output text as compared to the input text.

Abstract: Systems, computer-implemented methods, and computer program products that can facilitate predicting a source of a subsequent spoken dialogue are provided. According to an embodiment, a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a speech receiving component that can receive a spoken dialogue from a first entity. The computer executable components can further comprise a speech processing component that can employ a network that can concurrently process a transition type and a dialogue act of the spoken dialogue to predict a source of a subsequent spoken dialogue.

Abstract: A first text corpus comprising punctuated and capitalized text is received. The words in the first text corpus are then annotated with a set of labels indicating a punctuation and a capitalization of each word. At an initial training stage, a machine learning model is trained on a first training set using the annotated words from the first text corpus and the labels. A second text corpus is received representing conversational speech. The words in the second text corpus are then annotated with the set of labels. In a re-training stage, the machine learning model is re-trained on a second training set comprising the annotated words from the second text corpus, and the labels. At an inference stage, the trained machine learning model is applied to a target set of words representing conversational speech to predict a punctuation and capitalization of each word in the target set.

Abstract: In various embodiments, a quality inference application estimates perceived audio quality. The quality inference application computes a set of feature values for a set of audio features based on an audio clip. The quality inference application then uses a trained multitask learning model to generate predicted labels based on the set of feature values. The predicted labels specify metric values for metrics that are relevant to audio quality. Subsequently, the quality inference application computes an audio quality score for the audio clip based on the predicted labels.

Abstract: A computerized method is provided for automatically determining answers to a plurality of questions. The method includes automatically discovering a plurality of questions by processing historical data related to prior customer interactions. The automatically discovering includes applying a linguistic analytical model on the data related to historical customer interactions to detect the plurality of questions, vectoring the plurality of questions to generate mathematical representations of the questions, and grouping the plurality of questions into one or more clusters in accordance with similarities of the questions as measured based on their mathematical representations. The method also includes identifying the questions that do not have an existing answer. The method further includes determining at least one probable answer to each of the representative questions using a content mining technique that mines pertinent data from one or more identified content sources.

Abstract: An electronic apparatus configured to acquire information on a plurality of candidate texts corresponding to input speech of a user through a general speech recognition module, determine text corresponding to the input speech from among the plurality of candidate texts using a trained personal language model, and output the text as a result of speech recognition of the input speech.

Abstract: A recording device records a video and an imaging time, and a voice. Based on the voice, a sound parameter calculator calculates a sound parameter for specifying magnitude of the voice in a monitoring area at the imaging time for each of pixels and for each of certain times. A sound parameter storage unit stores the sound parameter. A sound parameter display controller superimposes a voice heat map on a captured image of the monitoring area and displays the superimposed image on a monitor. At this time, the sound parameter display controller displays the voice heat map based on a cumulative time value of magnitude of the voice, according to designation of a time range.

Abstract: A pattern recognition apparatus includes: a model storage part that stores a model(s) generated by using transfer path information indicating a difference of transfer paths of a signal(s) for training, additional to the signal(s) for training, and a pattern recognition part that inputs an input signal and transfer path information indicating a difference of transfer paths of the input signal, and performs pattern recognition of the input signal by using the model(s).

Abstract: A framework for reducing the number of textual items reviewed to determine the source of or reason for an anomaly in a time series that is used to track metrics in textual data is provided. According the framework, textual items in a time window corresponding to the anomaly are ranked according to the cross-entropy as determined by applying a language model to the relevant textual items and ranking textual items that most likely triggered an anomaly in time series data based on the cross-entropy value. In an aspect, a predetermined number of textual items having the highest cross-entropy are provided or all textual items having cross-entropy value higher than predetermine threshold are provided.

Abstract: A method may include receiving an electronic message from a sender. The method may further include parsing the electronic message into a set of sections, the set of sections including structured sections and an unstructured section. The method may further include detecting etiquette errors in the structured sections of the electronic message, wherein the etiquette errors include at least one of a missing word, a redundant word, an incorrect usage of a word, a style error, an emotional punctuation error, or a punctuation error. The method may further include generating an etiquette score based on the etiquette errors.

Abstract: A natural language processing system that trains task models for particular natural language tasks programmatically generates additional utterances for inclusion in the training set, based on the existing utterances in the training set and the existing state of a task model as generated from the original (non-augmented) training set. More specifically, the training augmentation module 220 identifies specific textual units of utterances and generates variants of the utterances based on those identified units. The identification is based on determined importances of the textual units to the output of the task model, as well as on task rules that correspond to the natural language task for which the task model is being generated. The generation of the additional utterances improves the quality of the task model without the expense of manual labeling of utterances for training set inclusion.

Abstract: A voice morphing apparatus having adjustable parameters is described. The disclosed system and method include a voice morphing apparatus that morphs input audio to mask a speaker's identity. Parameter adjustment uses evaluation of an objective function that is based on the input audio and output of the voice morphing apparatus. The voice morphing apparatus includes objectives that are based adversarially on speaker identification and positively on audio fidelity. Thus, the voice morphing apparatus is adjusted to reduce identifiability of speakers while maintaining fidelity of the morphed audio. The voice morphing apparatus may be used as part of an automatic speech recognition system.