Amazon Technologies, Inc.

Rafal Kuklinski

Michael Martin George

Rohit Prasad

Andrew Smith

Karthik Ramakrishnan

Christopher Schindler

Sumedha Arvind Kshirsagar

Thomas Edward Merritt

Roberto Barra Chicote

Adam Franciszek Nadolski

Thomas Schaaf

Manish Kumar Dalmia

Bartosz Putrycz

Remus Razvan Mois

Marco Nicolis

Derek Christopher Murman

Daniel Korzekwa

Roger Alix-Gaudreau

Wiktor Dolecki

Techniques for ensuring content output to a user conforms to a quality of the user's speech, even when a speechlet or skill ignores the speech's quality, are described. When a system receives speech, the system determines an indicator of the speech's quality (e.g., whispered, shouted, fast, slow, etc.) and persists the indicator in memory. When the system receives output content from a speechlet or skill, the system checks whether the output content is in conformity with the speech quality indicator. If the content conforms to the speech quality indicator, the system may cause the content to be output to the user without further manipulation. But, if the content does not conform to the speech quality indicator, the system may manipulate the content to render it in conformity with the speech quality indicator and output the manipulated content to the user.

Content output management based on speech quality

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.

Text-to-speech (TTS) processing

Systems, methods, and devices for generating text-to-speech output using previously captured speech are described. Spoken audio is obtained and undergoes speech processing to create text. The resulting text is stored with the spoken audio, with both the text and the spoken audio being associated with the individual that spoke the audio. Various spoken audio and corresponding text are stored over time to create a library of speech units. When the individual sends a text message to a recipient, the text message is processed to determine portions of text, and the portions of text are compared to the library of text associated with the individual. When text in the library is identified, the system selects the spoken audio units associated with the identified stored text. The selected spoken audio units are then used to generate output audio data corresponding to the original text message, with the output audio data being sent to a device of the message recipient.

Text-to-speech processing using previously speech processed data

Features are disclosed for recognizing inappropriate content in an output. The offensive content may be generated as a result of a speech processing error. A system may identify the inappropriate elements of a generated output and select among different appropriate alternatives. The system may be adjusted based on certain user characteristics. The system may be localized based on language and cultural features. The system may modify the generated output based on characteristics such as the tolerance threshold of known persons in the proximity of the system. The tolerance threshold may further be used to personalize and modify available content. Models used by the system may be further trained using input from a user.

Controlling offensive content in output

Growing community of inventors

Otomin, Poland

Rafal Kuklinski