Title:
METHOD OF GENERATING QUESTIONS BY COMBINING TRIPLE AND ENTITY TYPE IN KNOWLEDGE BASE
Document Type and Number:
WIPO Patent Application WO/2022/041294
Kind Code:
A1
Abstract:
A method of generating questions by combining a triple and an entity type in a knowledge base. An input of a neural network model is a word vector sequence representing a reconstructed triple, and an output, obtained by processing the word vector sequence, is a word vector sequence used to represent a question. The present method comprises: building a new triple on the basis of entity types corresponding to a head entity and a tail entity in a triple; using the GloVe word embedding method to obtain a representation of the new triple and a question corresponding to the new triple; using an attention mechanism-based gating mechanism recurrent neural network to encode the representation of the new triple, and outputting a representation of a triple incorporating contextual information; using the attention mechanism-based gating mechanism recurrent neural network to decode the representation of the triple, and outputting a representation of the question corresponding to the triple, thereby outputting the question; and replacing the entity types in the question output by the model with specific entities to obtain a new question. In the present method, the triple is combined with information of entity types corresponding to the head entity and the tail entity in the triple in a knowledge base and, by means of the attention mechanism-based neural network model, a question more grammatically fluent and more related to the input triple is obtained.
More Like This:
Inventors:
CAI YI (CN)
XU JINGYUN (CN)
XU JINGYUN (CN)
Application Number:
PCT/CN2020/112924
Publication Date:
March 03, 2022
Filing Date:
September 02, 2020
Export Citation:
Assignee:
UNIV SOUTH CHINA TECH (CN)
International Classes:
G06F40/166
Foreign References:
| CN110647620A | 2020-01-03 | |||
| CN111488440A | 2020-08-04 | |||
| CN111563146A | 2020-08-21 | |||
| CN111368528A | 2020-07-03 | |||
| US20200183963A1 | 2020-06-11 |
Other References:
SONG ZEHAN: "Multi-Source Co-Attention Networks for Composite Question Generation", MASTER THESIS, TIANJIN POLYTECHNIC UNIVERSITY, CN, 1 August 2020 (2020-08-01), CN , pages 1 - 75, XP055905924, ISSN: 1674-0246, DOI: 10.27461/d.cnki.gzjdx.2020.000454
Attorney, Agent or Firm:
FOSHAN JUNCHUANG INTELLECTUAL PROPERTY AGENCY FIRM (CN)
Download PDF: