Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
QUESTION ANSWERING SYSTEM TRAINING DEVICE AND COMPUTER PROGRAM THEREFOR
Document Type and Number:
WIPO Patent Application WO/2017/038657
Kind Code:
A1
Abstract:
[Problem] To provide a training device for a why-type question answering system with which learning data for a classifier is created efficiently while limiting manual intervention as much as possible. [Solution] A training device 62 includes: a question issuing unit 100, which issues to a question answering system 60 a question stored in a question/predicted answer storage unit 76; an answer candidate filtering unit 102, an answer candidate assessment unit 104, a learning data generation/labeling unit 106, and a learning data selection unit 108, which generate and add to a learning data storage unit 64 learning data for a ranking unit 122 of the why-type question answering system 60, said learning data being generated from pairs comprising the question and each of multiple answer candidates output together with a score from the question answering system 60; and an iteration control unit 110, which controls the question issuing unit 100, the answer candidate filtering unit 102, the answer candidate assessment unit 104, the learning data generation/labeling unit 106, and the learning data selection unit 108 so as to repeat the learning by the learning device, the issuing of a question, and the addition of the learning data, until an end condition is fulfilled.

Inventors:
OH JONGHOON (JP)
TORISAWA KENTARO (JP)
HASHIMOTO CHIKARA (JP)
IIDA RYU (JP)
TANAKA MASAHIRO (JP)
KLOETZER JULIEN (JP)
Application Number:
PCT/JP2016/074903
Publication Date:
March 09, 2017
Filing Date:
August 26, 2016
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
NAT INST INF & COMM TECH (JP)
International Classes:
G06F17/30; G06N5/02; G06N20/10
Foreign References:
JP2015011426A2015-01-19
Other References:
JONG-HOON OH ET AL.: "A Semi-Supervised Learning Approach to Why-Question Answering", PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI-16), 17 February 2016 (2016-02-17), pages 3022 - 3029, XP055374050, Retrieved from the Internet [retrieved on 20160907]
JONG-HOON OH ET AL.: "Why-Question Answering using Intra- and Inter-Sentential Causal Relations", PROCEEDINGS OF THE 51ST ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, 9 August 2013 (2013-08-09), pages 1733 - 1743, XP055309080, Retrieved from the Internet [retrieved on 20160907]
ICHIRO YAMADA ET AL.: "A Study on Extraction of Typical Sentence Structures Expressing Object Relations using a Semi-supervised Learning", FIT2008 (THE 7TH FORUM ON INFORMATION TECHNOLOGY) KOEN RONBUNSHU, vol. 2, 20 August 2008 (2008-08-20), pages 167 - 168, XP009509539
TATSUKUNI INOUE ET AL.: "Semi-Supervised Japanese Word Sense Disambiguation Based on Two-Stage Classification of Unlabeled Data and Ensemble Learning", JOURNAL OF NATURAL LANGUAGE PROCESSING, vol. 18, no. 3, 28 June 2011 (2011-06-28), pages 247 - 271, XP055374066
DAVID YAROWSKY: "Unsupervised Word Sense Disambiguation Rivaling Supervised Methods", ACL '95 PROCEEDINGS OF THE 33RD ANNUAL MEETING ON ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, 1995, pages 189 - 196, XP002257461, Retrieved from the Internet
See also references of EP 3346394A4
Attorney, Agent or Firm:
SHIMIZU, Satoshi (JP)
Download PDF: