Google Inc.

Vishaal Prasad

Ben David Goodrich

Daniel Duckworth

Chinnadhurai Sankar

Arvind Neelakantan

Semih Yavuz

Thomas Bugnon

Nicholas Johnson

Kevin Song Zhu

Arthur Gaudriot

Keith Wedelich

Julian Bill

George Huang

Sha Chang

Jacob Levine

Methods, systems, and media for determining and presenting information related to embedded sound recordings are provided. In some embodiments, the method comprises: receiving a content item; extracting a sound recording from the content item; generating a melody fingerprint of the extracted sound recording; determining whether the melody fingerprint of the extracted sound recording matches one of a plurality of clusters of similar sounding sound recordings in a reference database, wherein each cluster in the plurality of clusters of similar sounding sound recordings is associated with ownership information based on a plurality of ownership information associated with each of the sound recordings in the cluster; in response to determining that the melody fingerprint of the extracted sound recording matches a cluster of similar sounding sound recordings, retrieving ownership information associated with the cluster; mapping the ownership information to the sound recording extracted from the content item; and causing an action to be performed on the content item based on the mapped ownership information.

Methods, systems, and media for rights management of embedded sound recordings using composition clustering

Training and/or utilizing a single neural network model to generate, at each of a plurality of assistant turns of a dialog session between a user and an automated assistant, a corresponding automated assistant natural language response and/or a corresponding automated assistant action. For example, at a given assistant turn of a dialog session, both a corresponding natural language response and a corresponding action can be generated jointly and based directly on output generated using the single neural network model. The corresponding response and/or corresponding action can be generated based on processing, using the neural network model, dialog history and a plurality of discrete resources. For example, the neural network model can be used to generate a response and/or action on a token-by-token basis.

Generating automated assistant responses and/or actions directly from dialog history and resources

Growing community of inventors

Mountain View, CA, United States of America

Vishaal Prasad