Title:
LEARNING DEVICE, LEARNING METHOD, AND LEARNING PROGRAM
Document Type and Number:
WIPO Patent Application WO/2019/151506
Kind Code:
A1
Abstract:
This learning device (20) has: a hypothesis input unit (22) that receives input of an N-best hypothesis for learning in which voice recognition accuracy is already known, the hypothesis being inputted as a plurality of systems for learning of which the accuracy is already known; and a learning unit (23) that, when feature amounts of each of two hypotheses from among a plurality of hypotheses are applied, learns a model representing a neural network configured to be capable of assessing increases and decreases in the precision of the two systems.
Inventors:
OGAWA ATSUNORI (JP)
DELCROIX MARC (JP)
KARITA SHIGEKI (JP)
NAKATANI TOMOHIRO (JP)
DELCROIX MARC (JP)
KARITA SHIGEKI (JP)
NAKATANI TOMOHIRO (JP)
Application Number:
PCT/JP2019/003734
Publication Date:
August 08, 2019
Filing Date:
February 01, 2019
Export Citation:
Assignee:
NIPPON TELEGRAPH & TELEPHONE (JP)
International Classes:
G10L15/32; G10L15/06; G10L15/16
Foreign References:
| JP2011243147A | 2011-12-01 |
Other References:
IWATATE, MASAKAZU: "Japanese Dependency Parsing Using a Tournament Model", JOURNAL OF NATURAL LANGUAGE PROCESSING, vol. 15, no. 5, 10 October 2008 (2008-10-10), pages 169 - 185
SHIMAOKA SONSE: "Learning of word vector in the autoencoder", PROCEEDINGS OF THE 19TH ANNUAL MEETING OF THE ASSOCIATION FOR NATURAL LANGUAGE PROCESSING, pages 612 - 615
KUDO, TAKU: "Japanese Dependency Parsing using Relative Preference of Dependency", IPSJ JOURNAL, vol. 46, no. 4, April 2005 (2005-04-01), pages 1082 - 1092
HERBRICH, RALF ET AL.: "Learning Preference Relations for Information Retrieval", AAAI TECHNICAL REPORT, 1998, pages 83 - 86, XP055628611
OGAWA, ATSUNORI: "Rescoring of N-Best speech recognition hypotheses using an encoder-classifier model that performs one-to-one hypothesis comparison", PROCEEDINGS OF 2018 SPRING RESEARCH CONFERENCE OF THE ACOUSTICAL SOCIETY OF JAPAN, March 2018 (2018-03-01), pages 23 - 24
SHIMAOKA SONSE: "Learning of word vector in the autoencoder", PROCEEDINGS OF THE 19TH ANNUAL MEETING OF THE ASSOCIATION FOR NATURAL LANGUAGE PROCESSING, pages 612 - 615
KUDO, TAKU: "Japanese Dependency Parsing using Relative Preference of Dependency", IPSJ JOURNAL, vol. 46, no. 4, April 2005 (2005-04-01), pages 1082 - 1092
HERBRICH, RALF ET AL.: "Learning Preference Relations for Information Retrieval", AAAI TECHNICAL REPORT, 1998, pages 83 - 86, XP055628611
OGAWA, ATSUNORI: "Rescoring of N-Best speech recognition hypotheses using an encoder-classifier model that performs one-to-one hypothesis comparison", PROCEEDINGS OF 2018 SPRING RESEARCH CONFERENCE OF THE ACOUSTICAL SOCIETY OF JAPAN, March 2018 (2018-03-01), pages 23 - 24
Attorney, Agent or Firm:
SAKAI INTERNATIONAL PATENT OFFICE (JP)
Download PDF: