Pre-research Study based on Bibliometrics, Deep Learning Research for Aspect-Based Sentiment Analysis

Abstract

 Background: Massive publications on deep learning research for aspect-based sentiment analysis are challenging for interested researchers who want to research this area. Purpose: to provide an overview and comprehensive analysis on the research trend, which include the growth of publications, the most used keywords, the most popular publication sources to publish and find literature, the most cited publication, the most productive researcher, the most productive institution and country affiliation. Method: This study used a bibliometric method to analyze Scopus's indexed publications from 2014 (the year when the first publication was first indexed) to 2020. A total of 222 publications were analyzed and visualized using the VosViewer software. Result: In general, there is an increase in the number of publications from year to year. Keyword visualization shows keywords related to text-based processing, deep learning architectures, the research object and media, and the application of the method. The most popular sources to publish and to find literature are the “Lecture Notes in Computer Science” and the “Expert Systems with Applications''. The most cited publication is “Deep Learning for Aspect-Based Sentiment Analysis: A Comparative Review”, written by Do, Prasad (cited 81 times). The most productive researcher is Zhang Y from China. The most productive institution is Nanyang Technological University (6 publications), and China has the highest number of publications (76 documents). Conclusion: The bibliometric method can provide a conclusive and comprehensive preliminary overview of research trends for interested researchers who want to start research about deep learning for aspect-based sentiment analysis. 

 Keywords: Bibliometrics; Deep learning; Aspect-based sentiment analysis; VosViewer