Translator Disclaimer
9 September 2019 Synonym Extraction in Shipping Industry with Distant Supervision and Deep Neural Network
Yan Zhao, Qianwei Liu, Fanghuai Hu, Gang Lu, Tao Zhang
Author Affiliations +
Abstract

Zhao, Y.; Liu, Q.-W.; Hu F.-H.; Lu, G., and Zhang, T., 2019. Synonym extraction in shipping industry with distant supervision and deep neural network. In: Gong, D.; Zhu, H., and Liu, R. (eds.), Selected Topics in Coastal Research: Engineering, Industry, Economy, and Sustainable Development. Journal of Coastal Research, Special Issue No. 94, pp. 455–459. Coconut Creek (Florida), ISSN 0749-0208.

Current synonym extraction methods, including lexical pattern based methods, distributional similarity-based methods, and search engine based methods, usually need a lot of manual work, either in rule editing or in training corpus labeling. This paper tries to find synonyms in a “self-supervised” way and presents a synonym extraction framework based on distant supervision. The model used here is LSTM and CRF. We first analyze the nature of the synonym representations of the Nautical Navigation in text; then automatically generate labeled training samples by using structured knowledge (a domain dictionary) and some generic heuristic rules. Finally, we train some “LSTM + CRF” models and use them to extract synonyms in the shipping industry corpus. More than 100 thousand facts are successfully extracted, which contain 3,629 distinct pairs of synonyms; and the precision is 0.967.

©Coastal Education and Research Foundation, Inc. 2019
Yan Zhao, Qianwei Liu, Fanghuai Hu, Gang Lu, and Tao Zhang "Synonym Extraction in Shipping Industry with Distant Supervision and Deep Neural Network," Journal of Coastal Research 94(sp1), 455-459, (9 September 2019). https://doi.org/10.2112/SI94-090.1
Received: 6 February 2019; Accepted: 13 March 2019; Published: 9 September 2019
JOURNAL ARTICLE
5 PAGES


Share
SHARE
KEYWORDS
distant learning
LSTM
Shipping industry
Synonym extraction
RIGHTS & PERMISSIONS
Get copyright permission
Back to Top