Title:
PHRASE VECTOR LEARNING DEVICE, METHOD, AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2019/160097
Kind Code:
A1
Abstract:
According to the present invention, a vector appropriate for an arbitrarily defined phrase can be created. A lattice construction unit 212 constructs a lattice structure that comprises a link connecting adjacent word or phrase candidates, on the basis of a morpheme analysis result or a dependency analysis result of an inputted text. A first learning unit 213 learns, on the basis of the lattice structure, a neural network A for estimating, from the word or phrase candidates, peripheral word or phrase candidates thereof. A vector generation unit 214 acquires vectors corresponding to the word or phrase candidates from the neural network A, and uses the vectors as learning data. A second learning unit learns a neural network B for vectorizing the word or phrase candidates on the basis of the learning data.
Inventors:
SAITO ITSUMI (JP)
NISHIDA KYOSUKE (JP)
ASANO HISAKO (JP)
TOMITA JUNJI (JP)
NISHIDA KYOSUKE (JP)
ASANO HISAKO (JP)
TOMITA JUNJI (JP)
Application Number:
PCT/JP2019/005621
Publication Date:
August 22, 2019
Filing Date:
February 15, 2019
Export Citation:
Assignee:
NIPPON TELEGRAPH & TELEPHONE (JP)
International Classes:
G06F16/00; G06F40/20; G06N3/02
Foreign References:
| JP2017027526A | 2017-02-02 |
Other References:
ASATANI, KIMITAKA: "Understanding Citation-networks by Network Embedding", INFORMATION PROCESSING SOCIETY OF JAPAN SIG TECHNICAL REPORT, vol. 2016 -IC, no. 3, 24 February 2016 (2016-02-24), pages 1 - 8, ISSN: 2188-885X
ITO, RYO: "Lexicon Creation for Financial Sentiment Analysis using Network Embedding", THE JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE SPECIAL INTEREST GROUP ON FINANCIAL INFORMATICS, vol. 7, 10 March 2017 (2017-03-10), pages 70 - 76
MIZUGUCHI, GAI: "Learning Distributional Representations of Phrases Using LSTM", THE 31ST ANNUAL CONFERENCE OF THE JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE, 26 May 2017 (2017-05-26), XP055634591, Retrieved from the Internet
ITO, RYO: "Lexicon Creation for Financial Sentiment Analysis using Network Embedding", THE JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE SPECIAL INTEREST GROUP ON FINANCIAL INFORMATICS, vol. 7, 10 March 2017 (2017-03-10), pages 70 - 76
MIZUGUCHI, GAI: "Learning Distributional Representations of Phrases Using LSTM", THE 31ST ANNUAL CONFERENCE OF THE JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE, 26 May 2017 (2017-05-26), XP055634591, Retrieved from the Internet
Attorney, Agent or Firm:
TAIYO, NAKAJIMA & KATO (JP)
Download PDF: