US 12,334,061 B2
Conversation analysis system
Kiyoshi Sekine, Tokyo (JP)
Assigned to Interactive Solutions Corp., Tokyo (JP)
Filed by INTERACTIVE SOLUTIONS CORP., Tokyo (JP)
Filed on Dec. 11, 2023, as Appl. No. 18/536,220.
Application 18/536,220 is a continuation of application No. 17/614,375, granted, now 11,881,212, previously published as PCT/JP2021/001581, filed on Jan. 19, 2021.
Claims priority of application No. 2020-013067 (JP), filed on Jan. 29, 2020.
Prior Publication US 2024/0105168 A1, Mar. 28, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G10L 15/18 (2013.01); G10L 15/187 (2013.01); G10L 15/22 (2006.01)
CPC G10L 15/1815 (2013.01) [G10L 15/187 (2013.01); G10L 15/22 (2013.01)] 7 Claims
OG exemplary drawing
 
1. A conversation analysis system comprising:
a processor embedded in a computer, whereby the processor comprises:
a voice analysis unit (3) that analyzes content included in a conversation, wherein the voice analysis unit (3) uses the conversation as an input to convert the conversation into digital information, and stores the digital information as a voice-analyzed conversation in a storage unit (39) in a memory of the computer;
a topic ascertaining unit (5) that ascertains a topic of the conversation, wherein the topic ascertaining unit (5) compares each word in the conversation with each topic word(s) to obtain topic words in the conversation if any word in the conversation matches any of the each topic word(s), and ascertains a topic of the conversation, and wherein the each topic word(s) can be input into the system in advance or stored in a topic dictionary in the storage unit (39);
a topic information extraction unit (11) that extracts to topic-related information from the conversation which consists of terms related to the ascertained topic obtained from various information websites that are connected via the internet, wherein the terms related to the ascertained topic are stored as candidates of correction terms;
a topic-related correction term storage unit (7) that stores, for each topic, pronunciations to be voice-recognized or terms that are voice-recognized, and the candidates of correction terms related to the pronunciations to be voice-recognized or the terms that are voice-recognized; and
a conversation correction unit (9); wherein the conversation correction unit (9) uses the topic of the conversation ascertained by the topic ascertaining unit (5) to read the correction term candidates from the topic-related correction term storage unit (7) and correct the conversation analyzed by the voice analysis unit (3) by correcting any of any lettered word in the voice-analyzed conversation that is incorrect based on the ascertained topic with the candidates of correction terms; and
a memory storing a machine learning program to store and automatically update the candidates of correction terms.