公开(公告)号：US11238839B2

公开(公告)日：2022-02-01

申请号：US16575889

申请日：2019-09-19

Applicant: Spotify AB

Inventor： François Pachet ,  Pierre Roy ,  Mathieu Ramona ,  Tristan Jehan ,  Juan José Bosch Vicente

IPC: G10H1/00 ,  G06N20/00 ,  G06F16/635 ,  G06F16/632 ,  G06K9/62

Abstract: Methods, systems and computer program products are provided for identifying an audio stem. Audio stems (t1, . . . , tN) are stored on a stem database and songs (S1, . . . , SP) made with at least a subset of the plurality of the audio stems (t1, . . . , tN) are stored on a song database. At least partially composed song (S*) having a predetermined number of pre-selected stems (k) are received. In turn, a probability vector (or relevance value or ranking) is produced for each stem (t1, . . . , tN) to be complementary to the at least partially composed song (S*).

公开(公告)号：US20210312941A1

公开(公告)日：2021-10-07

申请号：US17202841

申请日：2021-03-16

Applicant: Spotify AB

Inventor： Juan José Bosch Vicente ,  François Pachet ,  Pierre Roy ,  Mathieu Ramona ,  Tristan Jehan

IPC: G10L25/51 ,  G06F16/632 ,  G06N3/04 ,  G06N3/08 ,  G10L25/30

Abstract: Methods, systems and computer program products are provided for determining acoustic feature vectors of query and target items in a first vector space, and mapping the acoustic feature vectors to a second vector space having a lower dimension. The distribution of vectors in the second vector space can then be used to identify items from the same songs, and/or items that are complementary. A mapping function is trained using a machine learning algorithm, such that complementary audio items are closer in the second vector space than the first, according to a given distance metric.

公开(公告)号：US10997986B2

公开(公告)日：2021-05-04

申请号：US16575926

申请日：2019-09-19

Applicant: Spotify AB

Inventor： Juan José Bosch Vicente ,  François Pachet ,  Pierre Roy ,  Mathieu Ramona ,  Tristan Jehan

IPC: G10L25/51 ,  G10L25/30 ,  G06N3/08 ,  G06N3/04 ,  G06F16/632

Abstract: Methods, systems and computer program products are provided for determining acoustic feature vectors of query and target items in a first vector space, and mapping the acoustic feature vectors to a second vector space having a lower dimension. The distribution of vectors in the second vector space can then be used to identify items from the same songs, and/or items that are complementary. A mapping function is trained using a machine learning algorithm, such that complementary audio items are closer in the second vector space than the first, according to a given distance metric.

公开(公告)号：US20230223037A1

公开(公告)日：2023-07-13

申请号：US18090228

申请日：2022-12-28

Applicant: Spotify AB

Inventor： Juan José Bosch Vicente ,  François Pachet ,  Pierre Roy ,  Mathieu Ramona ,  Tristan Jehan

IPC: G10L25/51 ,  G10L25/30 ,  G06N3/04 ,  G06F16/632 ,  G06N3/08

CPC classification number: G10L25/51 ,  G10L25/30 ,  G06N3/04 ,  G06F16/634 ,  G06N3/08

Abstract: Methods, systems and computer program products are provided for determining acoustic feature vectors of query and target items in a first vector space, and mapping the acoustic feature vectors to a second vector space having a lower dimension. The distribution of vectors in the second vector space can then be used to identify items from the same songs, and/or items that are complementary. A mapping function is trained using a machine learning algorithm, such that complementary audio items are closer in the second vector space than the first, according to a given distance metric.

公开(公告)号：US20210090590A1

公开(公告)日：2021-03-25

申请号：US16575926

申请日：2019-09-19

Applicant: Spotify AB

Inventor： Juan José Bosch Vicente ,  François Pachet ,  Pierre Roy ,  Mathieu Ramona ,  Tristan Jehan

IPC: G10L25/51 ,  G10L25/30 ,  G06N3/04 ,  G06N3/08 ,  G06F16/632

Abstract: Methods, systems and computer program products are provided for determining acoustic feature vectors of query and target items in a first vector space, and mapping the acoustic feature vectors to a second vector space having a lower dimension. The distribution of vectors in the second vector space can then be used to identify items from the same songs, and/or items that are complementary. A mapping function is trained using a machine learning algorithm, such that complementary audio items are closer in the second vector space than the first, according to a given distance metric.

公开(公告)号：US11568886B2

公开(公告)日：2023-01-31

申请号：US17202841

申请日：2021-03-16

Applicant: Spotify AB

Inventor： Juan José Bosch Vicente ,  François Pachet ,  Pierre Roy ,  Mathieu Ramona ,  Tristan Jehan

IPC: G10L25/51 ,  G06N3/04 ,  G06N3/08 ,  G10L25/30 ,  G06F16/632

Abstract: Methods, systems and computer program products are provided for determining acoustic feature vectors of query and target items in a first vector space, and mapping the acoustic feature vectors to a second vector space having a lower dimension. The distribution of vectors in the second vector space can then be used to identify items from the same songs, and/or items that are complementary. A mapping function is trained using a machine learning algorithm, such that complementary audio items are closer in the second vector space than the first, according to a given distance metric.

公开(公告)号：US20210090536A1

公开(公告)日：2021-03-25

申请号：US16575889

申请日：2019-09-19

Applicant: Spotify AB

Inventor： François Pachet ,  Pierre Roy ,  Mathieu Ramona ,  Tristan Jehan ,  Juan José Bosch Vicente

IPC: G10H1/00 ,  G06N20/00 ,  G06K9/62 ,  G06F16/635 ,  G06F16/632

Abstract: Methods, systems and computer program products are provided for identifying an audio stem. Audio stems (t1, . . . , tN) are stored on a stem database and songs (S1, . . . , SP) made with at least a subset of the plurality of the audio stems (t1, . . . , tN) are stored on a song database. At least partially composed song (S*) having a predetermined number of pre-selected stems (k) are received. In turn, a probability vector (or relevance value or ranking) is produced for each stem (t1, . . . , tN) to be complementary to the at least partially composed song (S*).