CAUSAL RELATIONSHIP RECOGNITION APPARATUS AND COMPUTER PROGRAM THEREFOR

WIPO Patent Application WO/2018/066445

A1

[Problem] To provide a causal relationship recognition apparatus that is capable of highly accurately recognizing a causal relationship expression irrespective of whether there is a keyword or not. [Solution] A causal relationship recognition apparatus 30 includes: a candidate vector generation unit 50 that receives a causal relationship candidate 32, and generates a candidate vector representing a word string constituting the candidate; a context vector generation unit 52 that generates a content vector representing a context in which a noun phrase of a cause and a noun phrase of a result of the causal relationship candidate emerge; a binary pattern vector generation unit 54 that generates a word vector representing background knowledge for determining presence of a causal relationship between the noun phrase included in the cause and the noun phrase included in the result; an answer vector generation unit 56; a related passage vector generation unit 58 concerning background knowledge; and a multi-column convolutional neural network 60 that has executed learning in advance to determine whether or not the causal relationship candidate is a causal relationship upon reception of the word vectors.

KRUENGKRAI CANASAI (JP)
HASHIMOTO CHIKARA (JP)
TORISAWA KENTARO (JP)
KLOETZER JULIEN (JP)
OH JONGHOON (JP)
TANAKA MASAHIRO (JP)

PCT/JP2017/035148

April 12, 2018

September 28, 2017

Click for automatic bibliography generation  

NAT INST INF & COMM TECH (JP)

G06F17/30; G06F40/20; G06N5/02

WO2014017023A1

2014-01-30

JP2007219947A	2007-08-30
JP2013175097A	2013-09-05

SHIMIZU, Satoshi (JP)

View/Download PDF      PDF Help

Previous Patent: DC-DC CONVERTER

Next Patent: ELECTRONIC DEVICE AND METHOD FOR CONTROL OF LANE STATE OF ELECTRONIC DEVICE