Sentiment Analysis

Elena Negrea-Busuioc, National University of Political Studies and Public Administration, Romania
elena.negrea@comunicare.ro

Sentiment analysis (or opinion mining) refers to the process of analysing digital texts to understand how language is used to evaluate, express opinions, attitudes, and emotions. By combining natural language processing, machine learning, and computational linguistics, sentiment analysis contributes to understanding opinion dynamics by measuring the sentiment or tone of communication in large datasets, such as social media posts or online reviews. Using text classifiers, sentiment analysis typically assigns a positive or a negative viewpoint to a text, predicting its polarity. This technique is widely used in market and communication research to gauge consumer satisfaction and analyse media coverage of various political, economic, and social topics.

Sentiment analysis can be conducted manually (expensive, time-consuming) or automatically using sentiment analysis tools (cost-effective, timesaving). There are two common types of automated sentiment analysis: dictionaries-based and supervised machine learning approaches. Despite its popularity, the efficiency, validity, and reliability of automated sentiment analysis in accurately determining sentiment has been questioned. The validation of sentiment measurements derived from automated sentiment analysis in communication and social sciences research has been under scrutiny. Studies have shown that automated sentiment analysis methods, whether using dictionaries or supervised machine learning, are generally less reliable and are always outperformed by manual (human) coding.

Automated sentiment analysis, despite its increasing sophistication, still faces significant challenges in accurately capturing the sentiment of opinion expressions. These challenges stem from linguistic phenomena such as irony, sarcasm, context-dependent expressions, as well as the inherent emotionality and subjectivity of language use. While automated sentiment analysis can provide useful insights, its validity for research purposes is limited, and, therefore, the results should be carefully reviewed and validated by human annotators to ensure accuracy and reliability.

Keywords: computational linguistics, sentiment, opinion expression

References:
Boukes, M., Van de Velde, B., Araujo, T., & Vliegenthart, R. (2020). What’s the tone? Easy doesn’t do it: Analyzing performance and agreement between off-the-shelf sentiment analysis tools. Communication Methods and Measures, 14(2), 83-104.
Liu, B. (2012). Sentiment analysis and opinion mining. Morgan & Claypool Publishers.
Van Atteveldt, W., Van der Velden, M. A., & Boukes, M. (2021). The validity of sentiment analysis: Comparing manual annotation, crowd-coding, dictionary approaches, and machine learning algorithms. Communication Methods and Measures, 15(2), 121-140.