Neural typing entities in Chinese-pedia | |
You, Yongjian1; Zhang, Shaohua1; Lou, Jiong1; Zhang, Xinsong1; Jia, Weijia1,2 | |
2018 | |
Conference Name | 2nd Asia Pacific Web and Web-Age Information Management Joint Conference on Web and Big Data, APWeb-WAIM 2018 |
Source Publication | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
Volume | 10987 LNCS |
Pages | 385-399 |
Conference Date | 7 23, 2018 - 7 25, 2018 |
Conference Place | Macau, China |
Author of Source | Springer Verlag |
Abstract | Typing entities in structured sources such as Wikipedia has been well studied to construct English knowledge bases automatically. However, there still remain two tough challenges in typing entities in Chinese-pedia. The first one is that structured information from Chinese-pedia cannot assign entities fine-grained types due to its inaccuracy and coarseness. The other challenge is the incompletion of Chinese-pedia, which means we can only use limited attribute fields to type entities. In this paper, we propose a novel Hierarchical Neural System (HNS) to infer fine-grained types for entities in Chinese-pedia. The HNS contains three main models which are hierarchical attention model, feature fusion model and hierarchical classification model. The hierarchical attention model extracts features from entity description based on a bi-LSTM network with hierarchical attention mechanism to break the limitation of inaccurate Chinese-pedia. To deal with the incompletion of Chinese-pedia, the feature fusion model is presented to obtain type features from multi-source such as descriptions, info-boxes, and categories. Through this model, we fuse all the features from different sources together and reduce the features to low-dimensional and dense vectors. Finally, the hierarchical classification model is designed to infer fine-grained types for entities in Chinese-pedia with features obtained from the other two models. The experiments illustrate that HNS outperforms the start-of-art work by 15.6% on f1-score. © Springer International Publishing AG, part of Springer Nature 2018. |
DOI | 10.1007/978-3-319-96890-2_32 |
Language | 英语 |
Fulltext Access | |
Citation statistics | |
Document Type | Conference paper |
Collection | University of Macau |
Affiliation | 1.Shanghai Jiao Tong University, Shanghai; 200240, China; 2.University of Macau, 999078, China |
Recommended Citation GB/T 7714 | You, Yongjian,Zhang, Shaohua,Lou, Jiong,et al. Neural typing entities in Chinese-pedia[C]//Springer Verlag,2018:385-399. |
Files in This Item: | There are no files associated with this item. |
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment