Abstract: Document retrieval through assertion analysis on entities and document fragments is disclosed. A document is received. Logical structures and entities are extracted from the document by parsing the document. For an entity in the extracted entities, an object representing the entity is created, an assertion made in the document associated with the entity is determined, and the assertion is linked to the object representing the entity. A logical structure from the extracted logical structures and content of the logical structure containing the assertion are identified and linked to the object representing the entity.

Abstract: A method, an electronic device, and computer readable medium is provided. The method includes identifying a frequency of each word that is present within a set of words. The method also includes deriving relatedness values for pairs of words. Each pair of words includes a first word and a second word in the set of words. Each relatedness value corresponds to a respective one of the pairs of words. Each relatedness value is based on the identified frequencies that the first word and the second word of the respective pair of words are present within the set of words. The method further includes generating a matrix representing the relatedness values. The method additionally includes generating a language model that represents relationships between the set of words included in the matrix.

Abstract: Aspects relate to apparatuses and methods for generating queries and transcribing video resumes. An exemplary apparatus includes at least a processor and a memory communicatively connected to the processor, the memory containing instructions configuring the processor to receive, from a posting generator, a plurality of posting inputs from a plurality of postings, receive a video resume from a user, generate a plurality of queries as a function of the video resume based on a plurality of posting categories, transcribe, as a function of the plurality of queries, a plurality of user inputs from the video resume, wherein the plurality of user inputs is related to attributes of a user, and classify the plurality of user inputs to the plurality of posting inputs to match the user to the plurality of postings.

Abstract: The subject matter of this specification can be embodied in, among other things, a method that includes receiving geographical information derived from a non-verbal user action associated with a first computing device. The non-verbal user action implies an interest of a user in a geographic location. The method also includes identifying a grammar associated with the geographic location using the derived geographical information and outputting a grammar indicator for use in selecting the identified grammar for voice recognition processing of vocal input from the user.

Abstract: Provided are a system and method for modifying a speech recognition result. The method includes: receiving, from a device, text output from an automatic speech recognition (ASR) model of the device; identifying at least one domain related to the received text; selecting, from among a plurality of text modification models included in the server, at least one text modification model corresponding to the identified at least one domain; and modifying the received text by using the selected at least one text modification model.

Abstract: A system configured to process speech commands may classify incoming audio as desired speech, undesired speech, or non-speech. Desired speech is speech that is from a same speaker as reference speech. The reference speech may be obtained from a configuration session or from a first portion of input speech that includes a wakeword. The reference speech may be encoded using a recurrent neural network (RNN) encoder to create a reference feature vector. The reference feature vector and incoming audio data may be processed by a trained neural network classifier to label the incoming audio data (for example, frame-by-frame) as to whether each frame is spoken by the same speaker as the reference speech. The labels may be passed to an automatic speech recognition (ASR) component which may allow the ASR component to focus its processing on the desired speech.

Abstract: Disclosed herein is a cooking system: a cooking apparatus configured to input and output a speech, transmit speech data corresponding to the speech, and cook food in a cooking chamber; a first server configured to perform communication with the cooking apparatus, when speech data is received from the cooking apparatus, perform speech recognition based on the received speech data, transmit response information to the speech recognition to the cooking apparatus, obtain a menu requested by the user based on the received speech data, and transmit a cooking time and a cooking temperature for the obtained menu to the cooking apparatus; and a second server configured to store information about at least one recipe for a plurality of menus, perform communication with the first server, and transmit the information about at least one recipe to the first server.

Abstract: A communication system includes a vehicle communication assembly connected to a vehicle, including a transmitter and a receiver, and configured to wirelessly communicate with a remote entity, and a processing device and a memory coupled to the processing device. The memory includes computer-executable instructions that, when executed by the processing device, cause the processing device to receive a wireless signal including a speech communication from the remote entity at the receiver, analyze the wireless signal by a speech recognition module to identify the speech communication, and recognize a known directive within the speech communication based on stored contextual information.

Abstract: Systems, methods, and computer-readable media for correcting transcriptions created through automatic speech recognition. A transcription of speech created using an automatic speech recognition system can be received. One or more domain-specific contexts associated with the speech can be identified and a text span that includes a mistranscribed entry can be recognized from the speech based on the one or more domain-specific contexts. Additionally, features can be extracted from the mistranscribed entry and the extracted features can be matched against an index of domain-specific entries to identify a correct entry of the mistranscribed entry. Subsequently, the transcription can be corrected by replacing with the mistranscribed entry with the correct entry.

Abstract: A method comprising detecting an activation of an intelligent assistant on an electronic device, waking up the intelligent assistant from a sleep mode in response to the activation, and determining an amount of vocabulary the intelligent assistant acts upon during a listening mode based on a type of the activation.

Abstract: The present invention is directed to a deep neural network (DNN) having a triplet network architecture, which is suitable to perform speaker recognition. In particular, the DNN includes three feed-forward neural networks, which are trained according to a batch process utilizing a cohort set of negative training samples. After each batch of training samples is processed, the DNN may be trained according to a loss function, e.g., utilizing a cosine measure of similarity between respective samples, along with positive and negative margins, to provide a robust representation of voiceprints.

Abstract: Systems and methods for automatically identifying problem-relevant sentences in a transcript are disclosed. In an example method, a transcript may be received of a first support call. A region of the first support call transcript may be identified, and first customer utterances may be detected in the region using a trained classification model. A trained regression model may estimate a relevancy to the problem statement of each of the first customer utterances, and one or more most problem-relevant statements may be selected from the first customer utterances, based on the estimated relevancies.

Abstract: A system capable of performing natural language understanding (NLU) on utterances including complex command structures such as sequential commands (e.g., multiple commands in a single utterance), conditional commands (e.g., commands that are only executed if a condition is satisfied), and/or repetitive commands (e.g., commands that are executed until a condition is satisfied). Audio data may be processed using automatic speech recognition (ASR) techniques to obtain text. The text may then be processed using machine learning models that are trained to parse text of incoming utterances. The models may identify complex utterance structures and may identify what command portions of an utterance go with what conditional statements. Machine learning models may also identify what data is needed to determine when the conditionals are true so the system may cause the commands to be executed (and stopped) at the appropriate times.

Abstract: In order to quickly respond to a question for vehicle information from a vehicle user at a remote location and offer the user a sense of relief, a vehicle management system of the present invention comprises a telematics communication unit which is mounted on a vehicle and acquires vehicle information and, a vehicle information server which receives from the telematics communication unit the vehicle information and a time at which the vehicle information is acquired, stores the received vehicle information and time, and transmits the stored vehicle information and time to a speech processing system as a response to a question when receiving the question about the vehicle information from the speech processing system.

Abstract: A computing system receives requests from client devices to process voice queries that have been detected in local environments of the client devices. The system identifies that a value that is based on a number of requests to process voice queries received by the system during a specified time interval satisfies one or more criteria. In response, the system triggers analysis of at least some of the requests received during the specified time interval to trigger analysis of at least some received requests to determine a set of requests that each identify a common voice query. The system can generate an electronic fingerprint that indicates a distinctive model of the common voice query. The fingerprint can then be used to detect an illegitimate voice query identified in a request from a client device at a later time.

Abstract: A method of training a custom hotword model includes receiving a first set of training audio samples. The method also includes generating, using a speech embedding model configured to receive the first set of training audio samples as input, a corresponding hotword embedding representative of a custom hotword for each training audio sample of the first set of training audio samples. The speech embedding model is pre-trained on a different set of training audio samples with a greater number of training audio samples than the first set of training audio samples. The method further includes training the custom hotword model to detect a presence of the custom hotword in audio data. The custom hotword model is configured to receive, as input, each corresponding hotword embedding and to classify, as output, each corresponding hotword embedding as corresponding to the custom hotword.

Abstract: A method, system, and computer program product for using a natural language processor includes importing a training text including a plurality of training nodes and having a first marked span that includes some of the plurality of training nodes and a first training trigger, wherein each training node includes a first attribute and a second attribute. Also included are generating a first plurality of training tree tables for each training node with respect to a first training traversal from the first training trigger, each training tree table including the first and second attributes of each training node in the first training traversal, and generating a first plurality of training comparison tables from the first plurality of training tree tables. Each training comparison table is also analyzed to determine if a target node belongs in the first marked span, and labeling each training comparison table accordingly.

Abstract: An apparatus for generating a score signal representing the quality of an audio or video signal supplied to the apparatus is proposed. The apparatus comprises: an input for supplying an audio or video signal, a computing unit implementing a neural network, the computing unit being supplied with the audio or video signal, and producing a score signal representing the quality of an audio or video signal supplied representing at least one predefined quality parameter of the audio or video signal, the neural network being set up by being trained with training data of a specific transmission standard and/or codec used for generating the audio or video data.

Abstract: A computer receives a first audio content item and applies a process to generate a representation of first audio content item. A portion is extracted from the representation of the first audio content item. A first representative vector that corresponds to the first audio content item is determined by applying a variational autoencoder (VAE) to a first segment of the extracted portion the audio content item. The computer stores the first representative vector that corresponds to the first audio content item.

Abstract: Identifying unreported issues in computing applications based on customer service interactions and website analytics. A computing system may receive communication data between a CSP and a user. The system may analyze the communication data using an NLP algorithm to identify a plurality of concepts in the communication data. The system may identify, based on the concepts, a target application associated with the communications between the CSP and the user. The system may receive analytics data from a web server hosting the target application and identify a feature of the target application that is not functioning. The system may then assign a priority to the feature of the target application that is not functioning based on a type of the feature and the received analytics data and generate a service issue record for the feature of the target application that is not functioning and the assigned priority.