Abstract: A target voice dataset may be augmented using speech prediction. Encoder and decoder models may be trained to encode audio data into encoded speech data and convert it back to audio. The encoded units may include semantic information (e.g., phonemes and/or words) as well as feature data indicating prosody, timbre, speaker identity, speech style, emotion, etc. of speech. An acoustic/semantic language model (ASLM) may be configured to predict encoded speech data in a manner analogous to a language model predicting words; for example, based on preceding encoded speech data. The models may be used to generate synthesized speech samples having voice characteristics (e.g., feature data) similar to those of the target voice dataset. The augmented dataset may be used to train a text-to-speech (TTS) model to reproduce the target voice characteristics, and may improve performance of the TTS model over training with only the original target voice dataset.

Abstract: A target voice dataset may be augmented using speech prediction. Encoder and decoder models may be trained to encode audio data into encoded speech data, and convert it back to audio. The encoded units may include semantic information (e.g., phonemes and/or words) as well as feature data indicating prosody, timbre, speaker identity, speech style, emotion, etc. of speech. An acoustic/semantic language model (ASLM) may be configured to predict encoded speech data in a manner analogous to a language model predicting words; for example, based on preceding encoded speech data. The models may be used to generate synthesized speech samples having voice characteristics (e.g., feature data) similar to those of the target voice dataset. The augmented dataset may be used to train a text-to-speech (TTS) model to reproduce the target voice characteristics, and may improve performance of the TTS model over training with only the original target voice dataset.

Abstract: A neural text-to-speech system may be configured to emphasize words. Applying emphasis where appropriate enables the TTS system to better reproduce prosodic characteristics of human speech. Emphasis may make the resulting synthesized speech more understandable and engaging than synthesized speech lacking emphasis. Emphasis may be manually annotated to, and/or predicted from, a source text (e.g., a book). In some implementations, the system may use a generative model such as a variational autoencoder to generate word acoustic embeddings indicating how emphasis is to be reflected in the synthesized speech. A phoneme encoder of the TTS system may process phonemes to generate phoneme embeddings. A decoder may process the word acoustic embeddings and the phoneme embeddings to generate spectrogram data representing the synthesized speech.

Abstract: Systems and methods for generating output audio with emphasized portions are described. Spoken audio is obtained and undergoes speech processing (e.g., ASR and optionally NLU) to create text. It may be determined that the resulting text includes a portion that should be emphasized (e.g., an interjection) using at least one of knowledge of an application run on a device that captured the spoken audio, prosodic analysis, and/or linguistic analysis. The portion of text to be emphasized may be tagged (e.g., using a Speech Synthesis Markup Language (SSML) tag). TTS processing is then performed on the tagged text to create output audio including an emphasized portion corresponding to the tagged portion of the text.

Abstract: During text-to-speech processing, audio data corresponding to a word part, word, or group of words is generated using a trained model and used by a unit selection engine to create output audio. The audio data is generated at least when an input word is unrecognized or when a cost of a unit selection is too high.

Abstract: Systems and methods for generating output audio with emphasized portions are described. Spoken audio is obtained and undergoes speech processing (e.g., ASR and optionally NLU) to create text. It may be determined that the resulting text includes a portion that should be emphasized (e.g., an interjection) using at least one of knowledge of an application run on a device that captured the spoken audio, prosodic analysis, and/or linguistic analysis. The portion of text to be emphasized may be tagged (e.g., using a Speech Synthesis Markup Language (SSML) tag). TTS processing is then performed on the tagged text to create output audio including an emphasized portion corresponding to the tagged portion of the text.

Abstract: Methods and devices for generating unique and different responses to commands are described herein. Natural language generation techniques may be employed to formulate responses to commands that are tailored to particular users. These responses account for previously provided responses, previously commands that have been made, and/or geographic locations of the requesting individual, for example. In some embodiments, an audible command may be received by a backend system from a voice activated electronic device. Text data may be generated from the audible command, and a user intent of the command is determined. Based on the user intent, a response from a particular application may be obtained. The response may be compared with previously generated responses and, if a similar responses was determined to have been provided previously, one or more different words, or a different arrangement of words, may be used to generate a new response.

Abstract: Systems and methods for generating output audio with emphasized portions are described. Spoken audio is obtained and undergoes speech processing (e.g., ASR and optionally NLU) to create text. It may be determined that the resulting text includes a portion that should be emphasized (e.g., an interjection) using at least one of knowledge of an application run on a device that captured the spoken audio, prosodic analysis, and/or linguistic analysis. The portion of text to be emphasized may be tagged (e.g., using a Speech Synthesis Markup Language (SSML) tag). TTS processing is then performed on the tagged text to create output audio including an emphasized portion corresponding to the tagged portion of the text.

Abstract: Methods and devices for generating unique and different responses to commands are described herein. Natural language generation techniques may be employed to formulate responses to commands that are tailored to particular users. These responses account for previously provided responses, previously commands that have been made, and/or geographic locations of the requesting individual, for example. In some embodiments, an audible command may be received by a backend system from a voice activated electronic device. Text data may be generated from the audible command, and a user intent of the command is determined. Based on the user intent, a response from a particular application may be obtained. The response may be compared with previously generated responses and, if a similar responses was determined to have been provided previously, one or more different words, or a different arrangement of words, may be used to generate a new response.