Research on Express Information Extraction Based on Multiple Sequence Labeling Models. Issue 2 (January 2021)

• Record Type: Journal Article
• Authors: Yu, Bihui; Wang, Ke
• Is Part Of: IOP conference series. Volume 632:Issue 2(2021)
• Publication Date: 2021-01
• Subjects: Information extraction -- Sequence annotation -- Bert -- Ernie; Earth sciences -- Periodicals; Environmental sciences -- Congresses; Environmental sciences -- Periodicals; 550.5
• DOI: 10.1088/1755-1315/632/2/022060
• Languages: English
• ISSNs: 1755-1307

Abstract

Abstract: With the popularity of online shopping, the logistics industry has developed rapidly. Major logistics companies have greater business needs to improve the efficiency of express sorting centers. Therefore, the technology of optimizing the efficiency of the express sorting system in the logistics industry and reducing the cost of time for customers to fill out express orders has received increasing attention from the society. The research work carried out in this paper is based on deep learning using natural language processing technology to extract express single text information, structured extraction of names and mobile phone numbers and addresses. The main content is that the express information structured extraction system automatically extracts information such as name, telephone, province, city, district, and detailed address. The docking of the express delivery sorting business system will help improve the logistics industry's operational capabilities. In this paper, four sets of sequence labeling models are constructed. Through multiple sets of data evaluation, it is shown that the model based on ERNIE and the learning rate decay strategy using Adam optimizer is superior to the current CNN, RNN, LSTM models. The highest F1 of the model constructed in this paper reaches 0.99473.