Abstract: A multi-material golf shaft having a butt portion joined to a tip portion and possessing unique relationships, including rigidity relationships, which provide beneficial performance characteristics.

Abstract: A health condition or change in health condition of a person may be determined by processing the person's speech with a neural network. Speech from more than one time period may be processed and, in some implementations, speech from a time period may be associated with a health condition label. For each time period, a feature vector may be computed from the speech and the feature vector may be processed with a neural network to obtain a speech embedding vector. In some implementations, feature vector may include word-piece encodings and the neural network may be a transformer neural network. The speech embedding vectors may be processed with a mathematical model to determine a change in a health condition between two time periods or to determine a health condition label for a specific time period. In some implementations, the mathematical model may be a regression model or a fully-connected neural network.

Abstract: A health condition or change in health condition of a person may be determined by processing the person's speech with a neural network. Speech from more than one time period may be processed and, in some implementations, speech from a time period may be associated with a health condition label. For each time period, a feature vector may be computed from the speech and the feature vector may be processed with a neural network to obtain a speech embedding vector. In some implementations, feature vector may include word-piece encodings and the neural network may be a transformer neural network. The speech embedding vectors may be processed with a mathematical model to determine a change in a health condition between two time periods or to determine a health condition label for a specific time period. In some implementations, the mathematical model may be a regression model or a fully-connected neural network.

Abstract: A health condition or change in health condition of a person may be determined by processing the person's speech with a neural network. Speech from more than one time period may be processed and, in some implementations, speech from a time period may be associated with a health condition label. For each time period, a feature vector may be computed from the speech and the feature vector may be processed with a neural network to obtain a speech embedding vector. In some implementations, feature vector may include word-piece encodings and the neural network may be a transformer neural network. The speech embedding vectors may be processed with a mathematical model to determine a change in a health condition between two time periods or to determine a health condition label for a specific time period. In some implementations, the mathematical model may be a regression model or a fully-connected neural network.

Abstract: A home interface comprises one or more audio interfaces within the premises of users. The audio interfaces are connected to an assistant service that detects conditions within rooms of user premises and that generates simulated audio themes for the rooms. The audio themes may be changed depending on detected conditions, and subthemes of a given audio theme may be implemented in different rooms.

Abstract: A user profile for a plurality of users may be built for speech recognition purposes and for acting as an agent of the user. In some embodiments, a speech processing device automatically receives an utterance from a user. The utterance may be analyzed using signal processing to identify data associated with the user. The utterance may also be analyzed using speech recognition to identify additional data associated with the user. The identified data may be stored in a profile of the user. Data in the user profile may be used to select an acoustic model and/or a language model for speech recognition or to take actions on behalf of the user.

Abstract: A device is configured with multiple applications that each respond to various commands. The correct application to receive a natural language command is identified by consideration of how well the command matches functions of the application. A target application to receive the command may additionally be selected by consideration of which application is most likely to receive a command. The likelihood of an application to receive a command may be determined by considering context. The command may be a voice input that is analyzed by speech recognition technology to determine word strings representing possible commands. Thus, the selection of a target application to receive the command may be based on any or all of the word strings from the natural language input, a closeness of fit between the command and an application, and the likelihood an application is the target for the next incoming command.

Abstract: A method for producing speech recognition results on a device includes receiving first speech recognition results, obtaining a language model, wherein the language model represents information stored on the device, and using the first speech recognition results and the language model to generate second speech recognition results.

Abstract: Techniques are described for determining locations of audio sources. Audio signals are captured from multiple locations. Pairs of the audio signals are analyzed to create correlograms, indicating correlation scores corresponding to different time offsets between the signals. Based on the correlograms, various locations are analyzed to determine probabilities of audio originating from those locations. The highest probabilities are found, indicating locations containing audio sources. In some situations, the audio sources may be reflective sources, and the locations of the reflective sources may be used to determine locations of objects or surfaces.

Abstract: A method for transcribing a spoken communication includes acts of receiving a spoken first communication from a first sender to a first recipient, obtaining information relating to a second communication, which is different from the first communication, from a second sender to a second recipient, using the obtained information to obtain a language model, and using the language model to transcribe the spoken first communication.

Abstract: A method for transcribing a spoken communication includes acts of receiving a spoken first communication from a first sender to a first recipient, obtaining information relating to a second communication, which is different from the first communication, from a second sender to a second recipient, using the obtained information to obtain a language model, and using the language model to transcribe the spoken first communication.

Abstract: A method for producing speech recognition results on a device includes receiving first speech recognition results, obtaining a language model, wherein the language model represents information stored on the device, and using the first speech recognition results and the language model to generate second speech recognition results.

Abstract: A method is given for automatically rewriting orthography of a stream of text words, for example, automatically and properly capitalizing words in the stream. If a word in the stream has an entry in an orthography rewrite lexicon, the word is automatically replaced with an orthographically rewritten form of the word from the orthography rewrite lexicon. In addition, selected words in the stream are compared to a plurality of features weighted by a maximum entropy-based algorithm, to automatically determine whether to rewrite orthography of any of the selected words.