Distant supervision for relation extraction uses an external knowledge base as supervision signals to automatically label corpus and has attracted more and more attention. However, this method has an ideal hypothesis that all instances containing the same entity pairs represent the same relation, which will lead to a lot of noisy data and affect the training effect of classifier. Aiming at the noise problem of distant supervised dataset, we propose a denoising method based on deep clustering. First, the deep clustering method is used to train the text features and clustering centers, and the effect of clustering is improved by using the information of mask entity pairs. Then high quality training samples are obtained by removing the noisy data, so as to achieve the effect of denoising. Experimental results show that the proposed model can effectively improve the performance of relation extraction by comparing with the traditional denoising methods.
A Denoising Method for Distant Supervised Relation Extraction Based on Deep Clustering
Lect. Notes Electrical Eng.
International Conference on Autonomous Unmanned Systems ; 2021 ; Changsha, China September 24, 2021 - September 26, 2021
Proceedings of 2021 International Conference on Autonomous Unmanned Systems (ICAUS 2021) ; Kapitel : 192 ; 1946-1956
2022-03-18
11 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Direct waveform extraction via a deep recurrent denoising autoencoder
British Library Conference Proceedings | 2019
|Semi-supervised classification based on clustering adjusted similarity
British Library Online Contents | 2017
|Wood inspection with non-supervised clustering
British Library Online Contents | 2003
|User-Interest-Based Document Filtering via Semi-supervised Clustering
British Library Conference Proceedings | 2005
|