Abstract: Implementations set forth herein relate to providing selectable autofill suggestions, which correspond to application actions that are at least partially fulfilled using server command data—prior to a user selecting a particular selectable autofill suggestion. Proactively fulfilling command data in this way mitigates latency between user selection of a suggestion and fulfillment of a particular action. Initially, a partial input can be processed to generate autofill suggestions, which can be communicated to a server device for further processing. The autofill suggestions can also be rendered for selection at a touch display interface, thereby allowing a user to select one of the autofill suggestions. As command fulfillment data is provided by the server, the command fulfillment data can be available to a corresponding application(s) in order that any corresponding actions can be at least partially fulfilled prior to user selection.

Abstract: Embodiments may provide improved techniques to assess model checkpoint stability on unseen data on-the-fly, so as to prevent unstable checkpoints from being saved, and to avoid or reduce the need for an expensive thorough evaluation. For example, a method may comprise passing a set of input sequences through a checkpoint of a sequence to sequence model in inference mode to obtain a set of generated sequences of feature vectors, determining whether each of a plurality of generated sequences of feature vectors is complete, counting a number of incomplete generated sequences of feature vectors among the plurality of generated sequences of feature vectors, generating a score indicating a stability of the model based on the count of incomplete generated sequences of feature vectors, and storing the model checkpoint when the score indicating the stability of the model is above a predetermined threshold.

Abstract: This disclosure describes, in part, techniques for utilizing global models to generate local models for electronic devices in an environment, and techniques for utilizing the global models and/or the local models to provide notifications that are based on anomalies detected within the environment. For instance, a remote system may receive an identifier associated with an electronic device and identify a global model using the identifier. The remote system may then receive data indicating state changes of the electronic device and use the data and the global model to generate a local model associated with the electronic device. Using the global model and/or local model, the remote system can identify anomalies associated with the electronic device and, in response to identifying an anomaly, notify the user. The remote system can further cause the electronic device to change states after receiving a request from the user.

Abstract: Processor(s) of a client device can: identify a textual segment stored locally at the client device; process the textual segment, using an on-device TTS generator model, to generate synthesized speech audio data that includes synthesized speech of the textual segment; process the synthesized speech, using an on-device ASR model to generate predicted ASR output; and generate a gradient based on comparing the predicted ASR output to ground truth output corresponding to the textual segment. Processor(s) of the client device can also: process the synthesized speech audio data using an on-device TTS generator model to make a prediction; and generate a gradient based on the prediction. In these implementations, the generated gradient(s) can be used to update weight(s) of the respective on-device model(s) and/or transmitted to a remote system for use in remote updating of respective global model(s). The updated weight(s) and/or the updated model(s) can be transmitted to client device(s).

Abstract: An acoustic model learning device is provided for obtaining an acoustic model used to synthesize voice signals with intonation. The device includes a first learning unit that learns the acoustic model to estimate synthetic acoustic feature values using voice and speaker determination models based on acoustic feature values of speakers, language feature values corresponding to the acoustic feature values and speaker data items, a second learning unit that learns the voice determination model to determine whether the synthetic acoustic feature value is a predetermined acoustic feature value or not based on the acoustic feature values and the synthetic acoustic feature values, and a third learning unit that learns the speaker determination model to determine whether the speaker of the synthetic acoustic feature value is a predetermined speaker or not based on the acoustic feature values and the synthetic acoustic feature values.

Abstract: A system includes at least one communication interface, at least one processor operatively connected to the at least one communication interface, and at least one memory operatively connected to the at least one processor and storing a plurality of natural language understanding (NLU) models. The at least one memory stores instructions that, when executed, cause the processor to receive first information associated with a user from an external electronic device associated with a user account, using the at least one communication interface, to select at least one of the plurality of NLU models, based on at least part of the first information, and to transmit the selected at least one NLU model to the external electronic device, using the at least one communication interface such that the external electronic device uses the selected at least one NLU model for natural language processing.

Abstract: The present invention provides a system for determining a language proficiency of a user in an evaluated language. A machine learning engine may be trained using audio file variables from a plurality of audio files and human generated scores for a comprehensibility, accentedness and intelligibility for each audio file. The system may receive an audio file from a user and determine a plurality of audio file variables from the audio file. The system may apply the audio file variables to the machine learning engine to determine a comprehensibility, an accentedness and an intelligibility score for the user. The system may determine one or more projects and/or classes for the user based on the user's comprehensibility score, accentedness score and/or intelligibility score.

Abstract: A computing device may receive audio data from a microphone representing audio in an environment of the device, which may correspond to an utterance and noise. A model may be trained to process the audio data to cancel noise from the audio data. The model may include an encoder that includes one or more dense layers, one or more recurrent layers, and a decoder that includes one or more dense layers.

Abstract: An approach is provided that receives a question at a question-answering (QA) system. A number of passages are identified that are relevant to the received question. A question knowledge graph is generated that corresponds to the question and a set of passage knowledge graphs are also generated with each passage knowledge graph corresponding to one of the identified passages. Each of the passage knowledge graphs are compared to the question knowledge graph with the comparison resulting in a set of knowledge graph candidate answers (kgCAs). A set of candidate answers (CAs) is computed by the QA with at least one of the CAs being based on one of the kgCAs.

Abstract: Voice synthesis method and apparatus generate second control data using an intermediate trained model with first input data including first control data designating phonetic identifiers, change the second control data in accordance with a first user instruction provided by a user, generate synthesis data representing frequency characteristics of a voice to be synthesized using a final trained model with final input data including the first control data and the changed second control data, and generate a voice signal based on the generated synthesis data.

Abstract: Conversations are modeled using dependency graph data structures to facilitate the interaction of users with automated assistants when performing actions performed by computing services. An automated assistant may utilize a dependency graph data structure to guide or otherwise control a human-to-computer dialog session with a user, e.g., by generating one or more outputs or prompts that are presented to the user on a computing device operated by that user, and may thereby enable efficient use of technical hardware.

Abstract: The present application discloses a method and an apparatus for training a speech spectrum generation model, as well as an electronic device, and relates to the technical field of speech synthesis and deep learning. A specific implementation is as follows: inputting a first text sequence into the speech spectrum generation model to generate an analog spectrum sequence corresponding to the first text sequence, and obtain a first loss value of the analog spectrum sequence according to a preset loss function; inputting the analog spectrum sequence corresponding to the first text sequence into an adversarial loss function model, which is a generative adversarial network model, to obtain a second loss value of the analog spectrum sequence; and training the speech spectrum generation model based on the first loss value and the second loss value.

Abstract: Provided is an artificial intelligence (AI) apparatus for generating a speech having a content-based style, including: a memory configured to store a plurality of TTS (Text-To-Speech) engines; and a processor configured to: obtain image data or text data containing a text, extract at least one content keyword corresponding to the text, determine a speech style based on the extracted content keyword, generate a speech corresponding to the text by using a TTS engine corresponding to the determined speech style among the plurality of TTS engines, and output the generated speech.

Abstract: This disclosure relates generally to system and method for hierarchical category classification of products. Generally in supervised hierarchical classification, the hierarchy structure is predefined. However, majority of the current machine learning methods either expect the model to learn the hierarchy from the data or requires separate models trained at each level taking the prediction of previous level as an additional input, thereby increasing latency in achieving training accuracy and/or requiring an explicit maintenance module to orchestrate inference and retrain multiple models (corresponding to the number of levels in the hierarchy). The disclosed method and system allows the predefined knowledge about hierarchy drive the learning process of a single model, which predicts all levels of the hierarchy. The disclosed multi-layer network model arrives at a consensus based on prediction at each level, thereby increasing the accuracy of prediction and reducing the training time.

Abstract: A method of generating diverse and natural text-to-speech (TTS) samples includes receiving a text and generating a speech sample based on the text using a TTS model. A training process trains the TTS model to generate the speech sample by receiving training samples. Each training sample includes a spectrogram and a training text corresponding to the spectrogram. For each training sample, the training process identifies speech units associated with the training text. For each speech unit, the training process generates a speech embedding, aligns the speech embedding with a portion of the spectrogram, extracts a latent feature from the aligned portion of the spectrogram, and assigns a quantized embedding to the latent feature. The training process generates the speech sample by decoding a concatenation of the speech embeddings and a quantized embeddings for the speech units associated with the training text corresponding to the spectrogram.

Abstract: Disclosed is speech synthesis in a noisy environment. According to an embodiment of the disclosure, a method of speech synthesis may generate a Lombard effect-applied synthesized speech using a feature vector generated from an utterance feature. According to the disclosure, the speech synthesis method and device may be related to artificial intelligence (AI) modules, unmanned aerial vehicles (UAVs), robots, augmented reality (AR) devices, virtual reality (VR) devices, and 5G service-related devices.

Abstract: Example data processing systems and methods are described. In one implementation, a system accesses a corpus of data and analyzes the data contained in the corpus of data to identify multiple documents. The system generates vector indexes for the multiple documents such that the vector indexes allow a computing system to quickly access the plurality of documents and identify an answer to a question associated with the corpus of data.

Abstract: Provided is pitch enhancement processing having little unnaturalness even in time segments for consonants, and having little unnaturalness to listeners caused by discontinuities even when time segments for consonants and other time segments switch frequently. A pitch emphasis apparatus carries out the following as the pitch enhancement processing: for a time segment in which a spectral envelope of a signal has been determined to be flat, obtaining an output signal for each of times in the time segment, the output signal being a signal including a signal obtained by adding (1) a signal obtained by multiplying the signal of a time, further in the past than the time by a number of samples T0 corresponding to a pitch period of the time segment, a pitch gain ?0 of the time segment, a predetermined constant B0, and a value greater than 0 and less than 1, to (2) the signal of the time.

Abstract: Methods, apparatuses, and systems of a common skill store are described herein to provide common skill storage and distribution for different virtual assistants on different provider platforms. A developer may not be bound by any virtual assistant or provider platform to develop skills in a required computer programming language, format, or style. They also need not to develop the same skill multiple times for different virtual assistants. After receiving a request to download a skill for use on any virtual assistant and provider platform, the common skill store may modify the requested program code to adapt it for the requesting virtual assistant. If a user has multiple user devices with different virtual assistants, they may only need to request the same skill one time, and the common skill store may send different sets of adapted program code to the multiple user devices with the different virtual assistants.

Abstract: An estimation device includes a memory, and processing circuitry coupled to the memory and configured to receive an input of an input audio signal that is an audio signal in which sounds from a plurality of sound sources are mixed, and an input of supplemental information, and output an estimation result of mask information that identifies a mask for extracting a sound of any one of the sound sources included in an entire or a part of a signal included in the input audio signal, the signal being identified by the supplemental information, cause a neural network to iterate a process of outputting the estimation result of the mask information, and cause the neural network to output an estimation result of the mask information for a different sound source, by inputting a different piece of the supplemental information to the neural network at each iteration.