Title: A named entity recognition method towards product reviews based on BiLSTM-attention-CRF

Authors: Shunxiang Zhang; Haiyang Zhu; Hanqing Xu; Guangli Zhu; Kuan-Ching Li

Addresses: School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan 232001, China ' School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan 232001, China ' School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan 232001, China ' School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan 232001, China ' Department of Computer Science and Information Engineering (CSIE), Providence University, Taichung 43301, Taiwan

Abstract: Named entity recognition (NER) towards product review intends to identify domain dependent named entities (e.g., organisation name, product name, etc.) from product reviews. Due to the fragmentation and non-construction of product reviews, traditional methods are difficult to capture the domain feature information and dependencies precisely. To solve the problem, we proposed a NER method towards product reviews based on BiLSTM-attention-CRF. Firstly, three kinds of features (character, word and part of speech) are integrated into the feature representation of texts. The final feature vector is obtained through training, mapping and linking the selected features. Then, the BiLSTM network is built to extract text features, and the attention mechanism is adopted to strengthen the capture of local features. Finally, CRF is applied to annotate and identify the entity. Compared with existing models, it is demonstrated that the proposed method can effectively recognise named entities from product reviews.

Keywords: named entity recognition; NER; product reviews; BiLSTM; attention; conditional random field; CRF.

DOI: 10.1504/IJCSE.2022.126251

International Journal of Computational Science and Engineering, 2022 Vol.25 No.5, pp.479 - 489

Received: 31 May 2021
Accepted: 01 Sep 2021
Published online: 18 Oct 2022 *

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