Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
MODEL LEARNING DEVICE, ESTIMATION DEVICE, METHOD THEREFOR, AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2018/147193
Kind Code:
A1
Abstract:
During model learning, a model learning device uses state-of-satisfaction variable pattern correct answer values, which represent correct answer values for a variable pattern of the state-of-satisfaction of a speaker in a conversation, and state-of-satisfaction correct answer values, which each represent correct answer values for the state-of-satisfaction of the speaker for each utterance made in a conversation, thereby obtaining, for each predetermined variable pattern for state-of-satisfaction, a state-of-satisfaction variable pattern model that contains a set of transition weights for a sequences of states of state-of-satisfaction; and uses learning-purpose utterance feature amounts and correct answer values for the state-of-satisfaction of the speaker making the learning-purpose utterances corresponding to the learning-purpose utterance feature amounts, thereby obtaining a state-of-satisfaction estimation model for obtaining a posterior probability of an utterance feature amount given a state-of-satisfaction of the speaker. During estimation, the model learning device uses an input utterance feature amount, the state-of-satisfaction variable pattern model, and the state-of-satisfaction estimation model, thereby obtaining an estimation value for the state-of-satisfaction of the speaker who made an utterance corresponding to the input utterance feature amount.

Inventors:
ANDO ATSUSHI (JP)
KAMIYAMA HOSANA (JP)
KOBASHIKAWA SATOSHI (JP)
Application Number:
PCT/JP2018/003644
Publication Date:
August 16, 2018
Filing Date:
February 02, 2018
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
NIPPON TELEGRAPH & TELEPHONE (JP)
International Classes:
G10L15/10; G06F40/20; G06Q10/04; G10L15/14
Domestic Patent References:
WO2014069076A12014-05-08
Foreign References:
US20100332287A12010-12-30
Other References:
ENGELBRECHT, K. P. ET AL.: "Modeling User Satisfaction with Hidden Markov Models", PROC. SIGDIAL, September 2009 (2009-09-01), pages 170 - 177, XP055533856
Attorney, Agent or Firm:
NAKAO, Naoki et al. (JP)
Download PDF: