US 12,141,207 B2
Major point extraction device, major point extraction method, and non-transitory computer readable recording medium
Setsuo Yamada, Tokyo (JP); Yoshiaki Noda, Tokyo (JP); and Takaaki Hasegawa, Tokyo (JP)
Assigned to NIPPON TELEGRAPH AND TELEPHONE CORPORATION, Tokyo (JP)
Appl. No. 17/268,471
Filed by NIPPON TELEGRAPH AND TELEPHONE CORPORATION, Tokyo (JP)
PCT Filed Aug. 14, 2019, PCT No. PCT/JP2019/031933
§ 371(c)(1), (2) Date Feb. 14, 2021,
PCT Pub. No. WO2020/036190, PCT Pub. Date Feb. 20, 2020.
Claims priority of application No. 2018-152891 (JP), filed on Aug. 15, 2018.
Prior Publication US 2021/0182342 A1, Jun. 17, 2021
Int. Cl. G10L 15/22 (2006.01); G06F 16/9032 (2019.01); G06F 16/9038 (2019.01); G06F 16/904 (2019.01)
CPC G06F 16/90332 (2019.01) [G06F 16/9038 (2019.01); G06F 16/904 (2019.01); G10L 15/22 (2013.01)] 16 Claims
OG exemplary drawing
 
1. A focus point extraction device comprising a computer configured to:
store a predetermined definition, the predetermined definition including (i) a first dialogue scene for which one or more utterance types are to be predicted, (ii) a second dialogue scene for which no utterance type is to be predicted, and (iii) an utterance content extraction method for each utterance type to be predicted for the first dialogue scene, the utterance content extraction method indicating which portion of an utterance belonging to the each utterance type is to be extracted as a focus point:
upon receipt of input of a dialogue including a plurality of utterances, predict dialogue scenes of the plurality of utterances;
predict, based on the predetermined definition, whether one or more utterances among the plurality of utterances whose dialogue scenes are predicted to correspond to the first dialogue scene belong to a prediction target utterance type, the prediction target utterance type being any utterance type to be predicted for the first dialogue scene; and
extract, in accordance with an utterance content extraction method for the prediction target utterance type included in the predetermined definition, focus points of respective utterances among the one or more utterances which are predicted to belong to the prediction target utterance type.