Title: Research on multi-feature fusion entity relation extraction based on deep learning

Authors: Shiao Xu; Shuihua Sun; Zhiyuan Zhang; Fan Xu

Addresses: School of Computer Science and Mathematics, Fujian University of Technology, 350118 Fuzhou, China ' School of Computer Science and Mathematics, Fujian University of Technology, 350118 Fuzhou, China ' School of Computer Science and Mathematics, Fujian University of Technology, 350118 Fuzhou, China ' School of Computer Science and Mathematics, Fujian University of Technology, 350118 Fuzhou, China

Abstract: Entity relation extraction aims to identify the semantic relation category between the target entity pairs in the original text and is one of the core technologies of tasks such as automatic document summarisation, automatic question answering system, and machine translation. Aiming at the problems in the existing relation extraction model that the local feature extraction of the text is insufficient and the semantic interaction information between the entities is easily ignored, this paper proposes a novel entity relationship extraction model. The model utilises a multi-window convolutional neural network (CNN) to capture multiple local features on the shortest dependency path (SDP) between entities, applies segmented bidirectional long short-term memory (BiLSTM) attention mechanism, extracts the global features in the original input sequence, and merges the local features with the global features to extract entity relations. The experimental results on the SemEval-2010 Task 8 dataset show that the model's entity relation extraction performance is further improved than existing methods.

Keywords: deep learning; multi-feature fusion; entity relation extraction; shortest dependency path; SDP; attention mechanism.

DOI: 10.1504/IJAHUC.2022.120949

International Journal of Ad Hoc and Ubiquitous Computing, 2022 Vol.39 No.1/2, pp.93 - 104

Received: 30 Jul 2020
Accepted: 04 Feb 2021
Published online: 18 Feb 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article