A Review of Feature Selection and Sentiment Analysis Technique in Issues of Propaganda

Abstract

Propaganda is a form of communication that is used in influencing communities, or people in general, to push forward an agenda for a certain goal. Nowadays, there are different means used in distributing propaganda including postings on social media, illustrations, cartoons and animations, articles, TV and radio shows. This paper is focused on election propaganda. Candidates in elections would use propaganda as a form of communication to channel and deliver messages through social media. Sentiment analysis (SA) is then used in identifying the positive and negative elements within the propaganda itself, through analysing the related documents, social media, articles or forums. This paper presents the various techniques used by previous researchers in issues of propaganda using SA, which include feature selection to remove irrelevant features and sentiment methods to identify sentiment in documents or others. Feature selection is a dominant side in sentiment analysis due to content of textual has a high measurement classification that can jeopardize SA classification interpretation. This paper also explores several SA techniques to identify sentiments in issues of propaganda. This study has also attempted to identify the use of swarm algorithms as a suitable feature selection method in SA for propaganda issues.

Sentiment analysis of tweets and government translations: Assessing China’s post-COVID-19 landscape for signs of withering or booming

“…Secondly, the study exclusively focuses on sentiment analysis of Twitter, which may be different from the broader population’s perspectives. Thirdly, the study does not consider the potential influence of censorship, propaganda, and social desirability bias on the sentiment analysis outcomes (Ahmad et al, 2019). Finally, the study’s reliance on sentiment analysis was conducted solely in English, which may not fully capture the nuances of sentiment expressed in other languages.

Figure 7.Timeline of average sentiments variations in 1 month.

…”

Proposing of suggestive influence detection and classification method based on fuzzy logic and feature driven analysis

“…Many techniques have been proposed to identify propaganda, disinformation, fake news and others information distortion techniques, which include data mining and text-mining using ensemble methods [1,2], linguistic-based detection and social network analysis methods [3], natural language inference approach [4], sentence-level analysis [5] and sentiment analysis techniques [6]. Due to generation of enormous amount of information pieces after full-scale Russian invasion binary classification methods using text-mining and dictionaries also regaining their popularity [7].…”

“…Let's summarize most important markers, relying on investigations [3,6,8,9,12,13,4,11,14], and create logically full list of notintersecting markers:…”

“…Analyzing existing feature-based approaches and relying on the feature selection rationale applicable to specifying information warfare tools, relying on the investigations [3,6,8,9,12,13,4,11,14] and stepping on the knowledge base on hybrid warfare tools provided by State Service of Special Communications and Information Protection of Ukraine and Center for Countering Disinformation the author proposed the correspondence of markers of suggestive influence to tools of informational influence.…”