Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speech recognition using neural networks. A feature vector that models audio characteristics of a portion of an utterance is received. Data indicative of latent variables of multivariate factor analysis is received. The feature vector and the data indicative of the latent variables is provided as input to a neural network. A candidate transcription for the utterance is determined based on at least an output of the neural network.

Abstract: The disclosed embodiments include computerized methods, systems, and devices, including computer programs encoded on a computer storage medium, for integrating voice-based interaction and control into a native graphical user interface (GUI) of an executed application. For example, a communications device may obtaining component data identifying a plurality of components of a voice-user interface from a computing system maintained by a voice-service provider, and may execute an application linked to a corresponding one of the components of the voice-user interface. The communications device may generate the native GUI based on an output of the executed application, and may generate an interface element representative of the corresponding one of the components of the voice-user interface. The communications device may present the generated interface element within the native GUI, which may embed the corresponding component of the voice-user interface into the native GUI.

Abstract: In some implementations, a device may monitor a set of data sources to generate a set of language models corresponding to the set of data sources. The device may determine a plurality of sets of keyword groups. The device may generate a plurality of sets of skill catalogs. The device may receive a source document for processing. The device may process the source document to extract a key phrase set and to determine a first similarity distance. The device may select a corresponding skill catalog and an associated language model based on a relevancy value. The device may determine second similarity distances between the source document and one or more target documents using the corresponding skill catalog and the associated language model. The device may output information associated with one or more target documents based at least in part on the second similarity distances.

Abstract: Methods and apparatus for selectively performing speech processing in a hybrid speech processing system. The hybrid speech processing system includes at least one mobile electronic device and a network-connected server remotely located from the at least one mobile electronic device. The mobile electronic device is configured to use an embedded speech recognizer to process at least a portion of input audio to produce recognized text. A controller on the mobile electronic device determines whether to send information from the mobile electronic device to the server for speech processing. The determination of whether to send the information is based, at least in part, on an analysis of the input audio, the recognized text, or a semantic category associated with the recognized text.

Abstract: Determining a language for speech recognition of a spoken utterance received via an automated assistant interface for interacting with an automated assistant. Implementations can enable multilingual interaction with the automated assistant, without necessitating a user explicitly designate a language to be utilized for each interaction. Implementations determine a user profile that corresponds to audio data that captures a spoken utterance, and utilize language(s), and optionally corresponding probabilities, assigned to the user profile in determining a language for speech recognition of the spoken utterance. Some implementations select only a subset of languages, assigned to the user profile, to utilize in speech recognition of a given spoken utterance of the user.

Abstract: Embodiments described herein provide a pipelined natural language question answering system that improves a BERT-based system. Specifically, the natural language question answering system uses a pipeline of neural networks each trained to perform a particular task. The context selection network identifies premium context from context for the question. The question type network identifies the natural language question as a yes, no, or span question and a yes or no answer to the natural language question when the question is a yes or no question. The span extraction model determines an answer span to the natural language question when the question is a span question.

Abstract: The present disclosure provides a modeling method for speech recognition and a device. The method includes: determining N types of tags; training a neural network according to speech data of Mandarin to generate a recognition model whose outputs are the N types of tags; inputting speech data of each dialect into the recognition model to obtain an output tag of each frame of the speech data of each dialect; determining, according to the output tags and tagged true tags, error rates of the N types of tags for the each dialect, generating M types of target tags according to tags with error rates greater than a preset threshold; and training an acoustic model according to third speech data of Mandarin and third speech data of the P dialects, outputs of the acoustic model being the N types of tags and the M types of target tags corresponding to each dialect.

Abstract: An electronic apparatus and a controlling method thereof are provided. The electronic apparatus includes a memory configured to store at least one instruction, and a processor configured to execute the at least one instruction to control the electronic apparatus to: determine a keyword from a query based on the query being input, obtain a word related to the keyword based on information on a user preference, and provide a response to the user query based on the keyword and the word. The processor may be configured to control the electronic apparatus to obtain at least one word from among a plurality of candidate words corresponding to the keyword as a word related to the keyword based on the user preference information. For example, at least part of a method of providing a response to a query by the electronic apparatus may use an AI model that is trained using at least one of machine learning, neural network or deep learning algorithm.

Abstract: According to embodiments of the present disclosure, a method, an apparatus, a device, and a computer-readable storage medium for determining a category of an entity are provided. The method includes: based on a suffix of the entity, obtaining a suffix feature associated with the suffix; determining one or more candidate categories of the entity based on a name of the entity; and determining a set of categories of the entity based on the one or more candidate categories and the suffix feature.

Abstract: A method of evaluating a language model using negative data may include accessing a first language model that is trained using a first training corpus, and accessing a second language model. The second language model may be configured to generate outputs that are less grammatical than outputs generated by the first language model. The method may also include training the second language model using a second training corpus, and generating output text from the second language model. The method may further include testing the first language model using the output text from the second language model.

Abstract: Knowledge gaps in a chatbot are identified with reference to a domain-specific document and a set of QA pairs of the chatbot. Entities and/or entity values associated with the document are compared to the entities and/or entity values of the QA pairs. Entities of the document not associated with the QA pairs are identified as knowledge gaps. The QA pairs and knowledge gaps are ranked by relevance to the domain.

Abstract: A method (and system) of determining confusable list items and resolving this confusion in a spoken dialog system includes receiving user input, processing the user input and determining if a list of items needs to be played back to the user, retrieving the list to be played back to the user, identifying acoustic confusions between items on the list, changing the items on the list as necessary to remove the acoustic confusions, and playing unambiguous list items back to the user.

Abstract: Disclosed herein are a real-time semantic annotation system and a method of converting user-entered natural language strings into semantically-readable knowledge structure documents using the system in real time. The real-time semantic annotation system includes a natural language character string input device for enabling a user to enter natural language character strings, a character string pattern triplet-mapping table for storing natural language character string patterns and their corresponding triplets, a triplet extraction device for converting the entered natural language character strings into triplets by analyzing and processing the entered natural language character strings using the pattern-triplet mapping table, an alternative word recommendation device for providing notification that a user should enter an alternative word, and a machine-readable document generation device for generating machine-readable documents from the triplets using a semantically-readable knowledge structure.

Abstract: Embodiments for converting a token collection that is derived from a natural language expression into a computational independent model (CIM) syntax tree representation are disclosed. In accordance with one embodiment, the conversion includes deriving a plurality of tokens from a natural language expression, where each of the plurality of tokens including at least one word. The conversion further includes transforming the plurality of tokens into a CIM syntax tree representation based on a CIM phrase tree model. The conversion also includes providing the CIM syntax tree representation to an application.

Abstract: A speech dialog system interfaces a user to a computer. The system includes a signal pre-processor that processes a speech input to generate an enhanced signal and an analysis signal. A speech recognition unit may generate a recognition result based on the enhanced signal. A control unit may manage an output unit or an external device based on the information within the analysis signal.

Abstract: A dictionary server includes a retrieval-display processing unit. Upon receipt of a request for retrieval of semantic information related to a term from a client PC, the retrieval-display processing unit acquires the semantic information, header information, and link information related to the semantic information from knowledge reference data, dictionary content data, and dictionary data. Based on the acquired information, the retrieval-display processing unit causes the client PC to display items on webpage related to the semantic information, the header information, and the link information.