	US 12,112,749 B2
	Command analysis device, command analysis method, and program
Kazunori Kobayashi, Tokyo (JP); Shoichiro Saito, Tokyo (JP); and Hiroaki Ito, Tokyo (JP)
Assigned to NIPPON TELEGRAPH AND TELEPHONE CORPORATION, Tokyo (JP)
Appl. No. 17/603,263
Filed by NIPPON TELEGRAPH AND TELEPHONE CORPORATION, Tokyo (JP)
PCT Filed Apr. 3, 2020, PCT No. PCT/JP2020/015261 § 371(c)(1), (2) Date Oct. 12, 2021, PCT Pub. No. WO2020/213427, PCT Pub. Date Oct. 22, 2020.
Claims priority of application No. 2019-078762 (JP), filed on Apr. 17, 2019.
Prior Publication US 2022/0199083 A1, Jun. 23, 2022
Int. Cl. G10L 15/00 (2013.01); G10L 15/04 (2013.01); G10L 15/05 (2013.01); G10L 15/22 (2006.01); G10L 25/87 (2013.01)

CPC G10L 15/22 (2013.01) [G10L 15/05 (2013.01); G10L 25/87 (2013.01); G10L 2015/223 (2013.01)]

3 Claims

1. A command analysis device, comprising circuitry configured to:

retrieve a final result recognition rule from a final result recognition rule storage device, a first intermediate result recognition rule from a first intermediate result recognition rule storage device, and a second intermediate result recognition result from the second intermediate recognition result storage device, wherein the first intermediate result recognition rule, the second intermediate result recognition rule, and the final result recognition rule are predetermined and are mutually distinct;

perform, every time when a predetermined unit of a speech signal is input, speech recognition on the speech signal and acquire a partial speech recognition result, which is a first intermediate result, wherein the perform operation further configured to:

determine whether the partial speech recognition result includes an end of a sentence, and output the speech recognition result, which is a final result, and an end-of-sentence signal when the end of a sentence is included;

determine, when the partial speech recognition result does not include the end of a sentence, whether a silence interval has continued for a predetermined period of time or more, or whether a new instance of the speech recognition result has not been acquired for a predetermined period of time or more, and output a silence signal when the silence interval has continued for the predetermined period of time or more or when a new instance of the speech recognition result has not been acquired for a predetermined period of time or more,

switch the first intermediate result recognition rule to the second intermediate result recognition rule upon receiving the silence signal, wherein the second intermediate result recognition rule allows commands to be analyzed even when the speech signal is interrupted at a point other than the end of a sentence, and

switch the first intermediate result recognition rule to the final result recognition rule, upon receiving the end-of-sentence signal, the final result recognition rule does not identify the command while the speech recognition is in progress;

verify the first intermediate result using the first intermediate result recognition rule to parse a command in the speech signal according to the speech recognition, wherein the first intermediate result recognition rule includes recognizing the command before the speech signal reaches an end of a sentence, wherein the verify operation is further configured to:

verify the intermediate result in light of the second intermediate result recognition rule when the intermediate result recognition rule is switched to the second intermediate result recognition rule, and

verify the final result in light of the final result recognition rule when the intermediate result recognition rule is switched to the final result recognition rule; and

output an analysis result during the input of the speech signal when a command execution target and a command execution content are successfully analyzed, wherein the analysis result indicates a type of the command conveyed by the speech signal.