Joint aspect-opinion extraction and sentiment orientation detection in university reviews

作     者：M, Devi Sri Nandhini Gurunathan, Pradeep 

作者机构：Department of Computer Science and Engineering A.V.C. College of Engineering Mannampandal Tamilnadu Mayiladuthurai India Department of Computer Applications A.V.C. College of Engineering Mannampandal Tamilnadu Mayiladuthurai India 

出 版 物：《International Journal of Information Technology (Singapore)》 (Int. J. Inf. Technol.)

年 卷 期：2022年第14卷第6期

页      面：3213-3225页

主　　题：Appositive adjectives Aspect-based sentiment analysis Aspect-sentiment word extraction Degree of intensification Implicit aspect extraction Nonary scale of sentiment strength Sentiment orientation detection 

摘      要：With the exponential growth of the Internet and the various forms of social media, the number of people using such platforms to share their views and experiences is also greatly increased. The reviews posted by people regarding a service or entity serve as valuable sources for decision making. However, acquiring valuable insights from the abundantly available unstructured information is not so straightforward. There is a need for more effective methodologies to carry out aspect-based sentiment analysis in a fine-grained manner. The proposed work performs joint aspect-opinion extraction and sentiment orientation detection (JAESOD) using an integrated approach of complex dependency rule-based extraction and proper treatment of various forms of adjectives that contribute to the effective extraction of subjective sentiment-bearing terms. The objective of the proposed work is three-fold: to extract aspect-sentiment word pairs, to detect implicit aspects, and to perform sentiment orientation detection at the aspect level along with sentiment scoring. In this paper, we propose a novel methodology to classify the sentiment orientation on a fine-tuned nine-point scale in contrast to the existing three to five-point scale found in the literature works. Concerning the task of sentiment detection, the goal of the proposed work is to achieve optimality in sentiment classification by giving due weightage to the modifies or intensifiers based on their degree of intensification and fair treatment of crucial factors such as double intensification, extended forms of adjectives, negations, sentiments expressed in numerical forms that play a vital role in fine-tuning the sentiment orientation detection. © 2022, The Author(s), under exclusive licence to Bharati Vidyapeeth s Institute of Computer Applications and Management.