Analysis of Peoples’ Opinion on Democratic Election in A Developing Nation Using Textblob Lexicon-Based Approach

Nigeria, one of the developing nations in Africa with over 200 million population, has a lot of capacity for economic growth, due to the large part of the 1958 discovery of significant oil reserves in the Niger Delta area of the country. Nigeria is endowed with a lot of natural resources that can boost its economic development, and many people both inside and outside the country believe that Nigeria would soon rise to occupy a major position in Africa and international affairs [1]. On October 1, 1960, Nigeria became independent from the British colonial administration. At that time, the future of the nation seemed bright and expectations were high. Throughout its history, Nigeria's system of government has undergone numerous changes. The nation switched to a parliamentary form of government after gaining independence, with the prime minister serving as the head of state. This system, however, was not long-lasting because several military takeovers in the 1960s and 1970s resulted in prolonged periods of military control. Nigeria did not revert to civilian administration until 1999, at which point a presidential system of government was put in place. Since then, the nation has chosen its presidents through routine elections which take place every four years.

Nigeria, the most populous nation in Black Africa and the self-described "giant of Africa," has improved its election system through a number of initiatives over colonial and post-colonial times, leading to electoral reforms in the country. The election management body (EMB) is in charge of supervising electoral administration in Nigeria and has undergone numerous changes or adjustments under different regimes to improve its election procedures. Nigeria's electoral reforms date back to the colonial era, when the elective principle was first incorporated into the political system in 1922 by the Clifford Constitution. Elections were held for four elective seats in the Nigerian Legislative Council, which was maintained and expanded by later colonial governments [2]. The Richards Constitution of 1946 created a central legislature for the nation, with three representatives from Lagos and one from Calabar, and four of the 24 members elected to the central legislature [3]. The first national election, with the exception of Lagos, was made possible by the Macpherson Constitution of 1951. Only in the northern region, where the election was indirect, was the first widespread direct election made possible by the Oliver Lyttleton Constitution of 1954 [4]. To conduct the 1959 elections and the nation's first direct elections, the departing colonial government established the Electoral Commission of Nigeria (ECN) in 1959 as a special electoral body [2]. Nigeria's electoral provisions order-in-council of 1958 stipulated that the governor general could designate an electoral commission from among neutral-minded individuals [5]. The Federal Electoral Commission (FEC), which was led by a Nigerian and held the country's first federal and regional elections in 1964 and 1965, was renamed the electoral umpire after Nigeria gained independence [6].

In 1976, General Olusegun Obasanjo's military government implemented a fresh election reform, replacing the FEC with the Federal election Commission (FEDECO). The 1979 and 1983 elections were held by FEDECO [7]. Also, the military government led by General Ibrahim Babangida instituted the National election Commission (NEC) in 1987 as part of yet another election reform. Elections for local government chairmen, lawmakers, governors, and presidents were held by NEC in 1990, 1991, and 1993, respectively. The 1995 establishment of the National Electoral Commission of Nigeria (NECON) by General Abacha's military government was another electoral reform initiative in Nigeria, spurred by the annulment of the presidential elections and other problems. The current Independent National Electoral Commission (INEC) was established in 1998 after General Abacha's death, when General Abdulsalami Abubakar dissolved the NECON. Seven elections have been held by INEC, the longest-serving EMB in Nigerian history: the 1999 transition election, the 2003 election, the 2007 election that resulted in the nation's first civilian regime change, the 2011 election [8], the 2015 election, the 2019 election, and the general elections that were just held in February 2023. Additionally, INEC has implemented a number of reforms over the years with the goal of strengthening democratic processes. The 1999 constitution (as modified), the Electoral Act of 2010 (as amended), and 2022 each govern elections in Nigeria.

Since 2011, there have been several modifications, such as the implementation of a permanent voter card, an electronic smart card reader, a simultaneous accreditation and voting system, and stakeholder participation prior to, during, and following all elections, including off-season ones [9]. Nigeria's democratic credentials were put to the test during the general election in 2023. While the governorship and state assembly elections were held on March 18, 2023, the presidential and national assembly elections were held on February 25, 2023. A year prior to the election, a new Electoral Act was passed, which authorised the electoral commission to use election technology, including electronic results transmission, and called for the early transfer of election funding to INEC. The Act enabled INEC to implement new technologies for the general election, including the INEC Election Results Viewing Portal (IReV) and the Bimodal Voter Accreditation System (BVAS). Interest in the election was raised by INEC's deployment of better technology for voter registration, particularly online pre-registration to vote. One positive aspect of the 2023 general election is the widespread implementation and high functionality rate of the BVAS for voter accreditation in the February 25 and March 18 elections.

The Nigerian presidential election is a highly anticipated moment with enormous political significance for the nation. Social media platforms like Facebook, Twitter, Instagram, and many others have significantly influenced how public opinions and political engagement have changed over the past years. A popular social media site called Twitter has emerged as an important tool for public dialogue, information exchange, and communication. It is impossible to ignore Twitter's impact in Nigeria. The website has developed into an essential forum for Nigerians, including activists, politicians, and citizens, to express their concerns, share news, inspire social movements, participate in political debates, and discuss pressing national issues by analyzing the language, tone, and context of tweets, voting patterns, and political preferences [10]. These platforms have developed into effective tools for political dialogue providing people with the opportunity to communicate their thoughts and feelings in an uncensored manner in the digital era. Twitter in particular has become well-known as a forum where people actively discuss politics, making it a useful source for sentiment research. One can learn a great deal about the opinions and attitudes of the public about different candidates, political parties, and important issues by examining Twitter feeds on the Nigerian presidential election. The majority of the decisions made in our daily lives are influenced by data, particularly data from social media sites like Twitter [10], and have been used to keep track of and predict elections. Twitter and other social media channels provide a wealth of real-time information that may be utilized to assess public opinion [11] and opinions of political candidates and parties [12]. Twitter users in particular constantly hunt for material to consume or share. Being a unique social networking site that makes it simple to start trends around topics by using hashtags rather than individual posts, which are prevalent on other social media sites. As a result, it is easier to keep track of users' opinions on particular subjects and their moods. During a study aimed at exploring the potential of the Twitter-sphere, researchers stumbled upon an unexpected finding. They discovered that political hashtags and the associated discussions exhibited the highest level of persistence among all the categories analyzed on Twitter [13].

The use of sentiment analysis is just one of the many methods that scholars have used to analyse opinions [14], others are natural language processing, and machine learning [15] to evaluate Twitter data and forecast the results of elections. Several techniques have been developed to track and forecast elections using information from Twitter [16]. Therefore, sentiment is an opinion, attitude, or assessment brought on by feeling. Sentiment analysis, commonly referred to as opinion mining, is a computational approach used to discern the emotional underpinnings of a text. It involves analyzing written language to identify and extract people's ideas, feelings, attitudes, and emotions expressed within the text [17]. Sentiment analysis has been used to reveal peoples’ opinion on elections in developing nations. Examples include the IndoBERT model and Twitter data that were used [18] to do a sentiment analysis of Indonesian public opinion on the 2024 election, and the study [19] computed sentiment scores by compiling tweets from major political parties that ran in the Gujarat Assembly Elections of 2022.

The primary purpose of sentiment analysis is to ascertain a speaker's or writer's viewpoint on certain topics or the overall contextual polarity of a work of literature. It establishes whether there are any positive, negative, or neutral feelings connected to a piece of writing. Sentiment analysis models are now frequently used in political methodology, starting with public political discourse [20]; this is consistent with the study of quantitative and qualitative methodologies used to understand politics and political systems [21]. Numerous studies have provided methods for comprehending politics that usually include formal theory, mathematics, and statistics. Election results have historically been anticipated analytically and statistically, with the technique depending on surveys and qualitative methodologies, researching political party manifestos, and keeping a watch on media trends [22].

The Nigerian presidential election of 2023 demonstrated the active involvement of individuals all around the nation, especially the lively youth population and knowledgeable elders including citizens who resided Overseas, this is due to their belief that they can change the bad economy of Nigeria through voting in the right candidate. Their participation, even on social media sites like Twitter, demonstrated their common ideas to tackle and ensure that Nigeria elects a qualified and deserving leader. However, there have been diverse expressions towards the outcomes of the general elections in Nigeria. Most of these views were expressed by Nigerians on cyberspace via Twitter, Facebook, Instagram, and so on. Therefore, there is a need to know the opinions of people towards the general election, to determine the most frequently used hashtags, and know the locations of the most active users. The 2023 Nigerian presidential election presents challenges in determining public opinion due to the diverse ideas of voters and the dynamic nature of sentiment. However, social media platforms like Twitter provide a valuable means for individuals to express their thoughts on candidates and the election results. Invariably, this work is focused on knowing the opinions of different categories of the populace regarding the outcome of the election. The scope of this study is to analyze tweets for positive, negative, or neutral sentiment for the 2023 Nigerian presidential election. Twitter served as the main data source for this project, which helped in collecting and analyzing a sizable number of tweets about the election. Focusing on voter sentiment, public opinion, and the connections between sentiment and election-related variables, the work investigates the sentiment trends, patterns, and correlations. By contrasting the results of the sentiment analysis with other measures of public opinion, the study evaluates the findings and debates their implications for the Nigerian presidential election in 2023. To clearly and understandably convey the sentiment analysis results, the project made use of data visualization approaches.

1.1 Adoption of new technologies versus traditional methods of election

Over the past 10 years, political campaigning has moved from offline—which includes traditional media and physical locations—to online. A model that has been drastically expanded upon and changed subsequently, the 2008 Obama campaign is sometimes credited as being the first to use social media mobilisation for electoral purposes. Additionally, because social media platforms have broken down some of the obstacles to engaging in democratic debate, some groups of people now have a comparatively easier time participating in conversations about elections and candidate accountability. First of all, it is important to acknowledge that social media platforms have upended the dominance of traditional media organisations, which has helped a lot of people by allowing new groups to participate in the democratic debate who were previously shut out of those structures. From the perspective of an election campaign, this has helped independent candidates and minor parties who didn't have the funds to run campaigns using conventional methods. Despite these benefits, it must be recognised that some groups frequently map onto preexisting social, economic, and political hierarchies, and that not everyone is similarly situated in digital domains.

People assume that everyone has access to the internet or digital gadgets since election debate and campaigning have moved online. This is undoubtedly not the case, especially in the developing world, where access to the internet is still very limited and varies by class, race/ethnicity, gender, (dis)abilities, language, digital skill levels, and occasionally caste and religion. The International Telecommunication Union (ITU) estimated that 2.7 billion people worldwide were not connected in 2022, which is about half of the world's population, according to the UN [23]; the majority of these individuals are in developing nations [24]. Individuals without access to the internet continue to experience "a worsening cycle of disenfranchisement" as digitisation and digital changes expand annually. There is still a "Gender Digital Divide" in the world, with 69% of men and 63% of women using the internet [24]. The disparity is considerably more pronounced in "least developed countries," where it stands at 13%. The number of women who own a mobile phone is 131 million lower than that of men [23]. Furthermore, the "gender gap in meaningful digital use" is even more pronounced, as studies show that women use digital devices less frequently due to a lack of technical skills and, in certain countries, familial pressures, which affects their capacity to engage in online politics. Political engagement is affected by these disparities, which also affect who has access to crucial voting information. Even if social media "creates value by attracting the attention of mainstream media," it has the effect of only reaching an elite audience in nations with low digital density. This feature is likely to impede efforts to enhance grassroots organizing [25].

The direct use of technology to conduct elections is worth examining individually, even though social media platforms' influence on electoral politics is multifaceted. Election management technologies have been used by states, political parties, and civil society. First of all, governments and political parties use technology extensively to raise awareness and register voters. While there may be some overlap with social media campaigning and influence strategies, election management explicitly focusses on the use of technology to disseminate information about election procedures. Voter registration lists, voting station locations, polling timetables, and polling results are all examples of this. To increase voter turnout and enhance voter registration, governments worldwide have looked to information, communication, and technology (ICT). Social media and electronic platforms have been utilised by Election Management Board (EMBs), civil society, and governments in general to broaden voter education and outreach. Social media in particular has been utilised in several nations to raise voter education among young voters, many of whom are first-time voters [26]. To raise awareness and help voters register, EMBs have used SMS services in other nations with limited internet and social media access. It is good to say that; the world is evolving and the use of ICT is one of the major paradigms bringing this change on a daily basis. Therefore, the adoption of ICT in electoral processes need to be embraced for easier and fast processing of electoral procedures.

This study used model research methods which hinged on quantitative and qualitative analysis of the data. To achieve the objectives of this research, four experimental procedures were followed and they were: data collection, data preprocessing, analysis of data, and data visualization. Figure 1 shows the research process flow, which is a step by step guide on how the research was done. Four distinct stages were identified for the realization of the research objectives.

1.png

Figure 1. Research process flow

3.1 Method

This study utilised Python 3.8 and TextBlob version 0.17.1. Python 3.8 and TextBlob version 0.17.1 can be a powerful combo for political sentiment analysis, mainly because of TextBlob's compatibility and capabilities with this Python version. Python 3.8 environments may be relied upon thanks to version 0.17.1's stability and strong support. TextBlob uses NLTK to give users access to other capabilities that are essential for sentiment analysis, like tokenisation, stopword elimination, and POS tagging. It easily interfaces with other libraries, such as pandas, making it possible to preprocess and analyse massive text datasets more easily. Even people with no prior knowledge of natural language processing (NLP) can use TextBlob's simple API to process textual input. It has a built-in sentiment analysis function that applies a rule-based methodology based on models that have already been trained. TextBlob is a great tool for testing concepts or small-scale projects because it allows for quick prototyping of sentiment analysis pipelines. You can examine public opinion on political candidates, policies, or events using TextBlob's sentiment score feature, and you can monitor changes in sentiment during elections, debates, and emergencies. Since propaganda and false information are common in political discourse, it is essential to recognise if a piece is subjective or factual.

The predefined sentiment lexicons that TextBlob uses are essentially databases of words with corresponding sentiment scores. The dictionary contains words with subjectivity (a value ranging from 0 (objective) to 1 (subjective) and polarity (a value ranging from -1 (negative sentiment) to +1 (positive sentiment). TextBlob automatically does sentiment analysis using the PatternAnalyzer, which is based on the vocabulary of the Pattern library. After tokenising the input into discrete words or phrases, TextBlob retrieves the polarity and subjectivity values of the tokens by comparing them with entries in the sentiment lexicon. The following natural language processing (NLP) tools and techniques are used: TextBlob, CountVectorizer, Truncated SVD, Latent Semantic Analysis (LSA), and Sentiment Intensity Analyser (e.g., VADER). When used together, they can enhance processes like political sentiment analysis.

3.2 Data collection

The dataset used in this analysis came from Twitter and included 85,662 tweets about the 2023 presidential election, since there is a need for a more accurate dataset, Twitter data from November 2022 to April 2023 had to be gathered, which indicated the time at which the presidential aspirants were made known to Nigerians and the commencement of discussions about the 2023 presidential elections on Twitter and other social media platforms. The Twitter API was used to stream real-time data alongside the Tweepy package in the Python programming language. Tweepy library is a package in Python that allow for real-time streaming of tweets via Twitter API. Since the analysis being carried out requires a real-time assessment, this account for the time frame which coincide with the election period in Nigeria. Each tweet was collected based on many factors like user_name, user_location, user_ description, user_description, user_created, etc., and the tweets made use of several trending hashtags like #NigeriaDecides2023, #NigeriaPresidentialElection2023, and many more, Nigerian Presidential candidates and tweets by citizens with top electoral preferences. The selection of the hashtag and keywords used for this study was based on the “trending” on Twitter App, individual hashtag common among the Nigeria citizens on their social media handles, and more importantly, selected keywords during the political talks aired on Arise TV, channel Television, TVC Political talks and the likes. A Twitter hashtag connects the conversations of several users into a single stream. This enables unrelated Twitter users to discuss the same subject. Their tweets show in the same stream because of the hashtag they employ. This resolves a coordination issue and helps to start a discourse. Since so many individuals are discussing the subject, some hashtag terms or phrases start to gain popularity. These are referred to as "trending topics."

3.3 Data preprocessing

This stage involves the cleaning of the dataset to reduce noise and increase the effectiveness of text analysis algorithms. The following techniques were used to clean up the tweets:

i. Text processing requires the removal of Unicode strings because some Unicode characters may not be supported by all systems or software, which might result in problems with text encoding and display.

ii. Substitution of blank lines for Re-tweets. In text processing, it's critical to remove the "RT" (re-tweet) prefix from text since it helps to avoid information duplication and lower noise levels in the text data. The effectiveness and precision of text processing algorithms can be increased by removing them and substituting blank spaces.

iii. Change any link to "URL" In text processing, it's crucial to convert URLs (Uniform Resource Locators) to the token "URL" because doing so helps to cut down on background noise in the text data. The main goal of this removal is to get rid of extraneous data and noise that could negatively impact sentiment categorisation systems' effectiveness. Frequently, URLs don't add any significant sentiment-related information to the textual material. If not managed well, they might cause needless complexity and misclassification. The textual information is made more concise by eliminating URLs, which frees up the sentiment analysis algorithms to concentrate on the linguistic content that actually expresses sentiment. Additionally, special characters like symbols and punctuation can obstruct text analysis. Exclamation marks and other punctuation can convey emotion, but too many or unnecessary special characters can add noise. Removing superfluous special characters aids in text data normalisation, which improves feature extraction and tokenisation. This, in turn, reduces dimensionality and focusses on important features, improving the performance of machine learning models used in sentiment analysis. For text data to be cleaned, noise levels to be decreased, and sentiment analysis models to be more accurate, URLs and special characters must be eliminated.

iv. It is necessary to change any instances of "@username" to "user" because it helps to clean up text data by lowering noise. On social media sites like Twitter, "@username" is the symbol for mentioning or tagging other users. The findings may be skewed by these mentions, which are generally irrelevant to text-processing tasks.

v. Eliminate more white spaces since doing so helps text processing algorithms and machine learning models work more effectively and accurately. The performance of text processing algorithms can be hampered by excessive white spaces, which can also affect text encoding and display issues.

vi. Removal of hashtags placed in front of words. On social networking sites like Twitter, hashtags are used to organize and label tweets, however, they are not always applicable to the given job. By lowering the dimensionality of the data, removing hashtags can aid in enhancing the effectiveness and precision of text processing algorithms and machine learning models. Understanding the purpose of the hashtags is essential for topic model and topic classification. Replacing hashtags with their titles is a simple method of incorporating this information. This can be accomplished by processing the HTML response after performing an HTTP GET.

vii. Remove digits, multiple exclamation points, and numerous question marks because they can skew and bring noise to text data. Numbers may not be relevant to the work at hand; several exclamation or question marks may show feeling, but they may also signify a mistake or a joke rather than a genuine feeling. By removing them, the dimensionality of the data may decrease, which assisted the text processing algorithms and machine learning models to work more effectively and accurately. In a situation where punctations are useful in the analysis, feature engineering on the occurrences of various punctuation, particularly on consecutive appearances should be done.

viii. Emotions (such as emoticons and emojis) should be removed from text data while processing it since they can contribute to noise and bias. Though they might also be unrelated to the job at hand and add bias in text categorization, sentiment analysis, and other NLP tasks, emotions are frequently utilized to express sentiment. For emoticons that have impact in the analysis, it should be converted to the English word.

ix. Lowercase is now used. Because it increases the effectiveness and precision of text processing algorithms and machine learning models, text to lowercase conversion in text processing is crucial. This is because a lot of machine learning and text processing algorithms are case-sensitive, and changing the text to lowercase can assist in making the data less dimensional and more consistent [54]. In case where the excitement is expressed using letter case, feature engineering needs to be done before conversion to lower case to preserve the meaning of the tweets.

3.4 Analysis of data

Text processing involves the crucial task of categorizing tweets as positive, negative, or neutral, accomplished through a text processing technique known as Count Vectorizer.

Count Vectorizer is a specific type of bag-of-words model that converts text into a numerical format. The CountVectorizer technique operates by initially tokenizing the text, breaking it down into individual words or phrases [44]. These tokens are subsequently tallied, and the counts are stored in a vocabulary. The vocabulary is a mapping from each token to its corresponding count. CountVectorizer in sentiment analysis is used to identify sentiment-related words. For example, words like "love" and "happy" are often associated with positive sentiment, while words like "hate" and "sad" are often associated with negative sentiment. CountVectorizer can be used to identify these words and then use them to create a sentiment lexicon. This lexicon can then be used to manually classify tweets or to train a machine learning model. Finally, CountVectorizer can also be used to analyze the sentiment of tweets over time. This can be done by tracking the frequency of sentiment-related words in tweets over time. This information can then be used to identify trends in public opinion.

Truncated SVD (Singular Value Decomposition) is a method for reducing the number of features in a dataset for dimensionality reduction. It works by performing a singular value decomposition (SVD) on the dataset. SVD decomposes the dataset into a set of orthogonal vectors, called singular vectors, and a set of singular values. The singular values are arranged in decreasing order, and the Truncated SVD algorithm keeps only the top $k$ singular values. TruncatedSVD is made use of in sentiment analysis to identify sentiment-related features. For example, Truncated SVD was used in this study when trying to identify the top 15 tweets that are most correlated with negative sentiments.

Latent Semantic Analysis (LSA) is a method for drawing out semantic data from text. The number of features in a dataset can be decreased using this kind of dimensionality reduction method. To use LSA, you must first create a term-document matrix. This matrix is a table that shows how often each term appears in each document. LSA then performs a singular value decomposition (SVD) on the term-document matrix. SVD decomposes the matrix into a set of orthogonal vectors, called singular vectors, and a set of singular values. The singular values are arranged in decreasing order, and the LSA algorithm keeps only the top $k$ singular values. The LSA algorithm then reconstructs the term-document matrix using the top k singular vectors. This results in a new term-document matrix with a lower dimensionality. Latent Semantic Analysis is made use of in sentiment analysis by implementing it with Truncated SVD in the dataset of the tweet of the 2023 presidential election to identify the top 10 topics that were prevalent during the elections.

Sentiment Intensity Analyser is a sentiment analysis tool that is made specifically to operate with social media text and is lexicon-based and rule-based. It works by first identifying sentiment-related words in a text. These words are then scored according to their intensity by getting a sentiment score between -1 and 1. A score of -1 represents a very negative sentiment, while a score of 1 represents a very positive sentiment. For example, the tweet "I'm so excited to vote for my candidate in the 2023 election!" would be scored as very positive, while the tweet "I'm so disappointed in the candidates in the 2023 election!" would be scored as very negative. Another example is the word "love" which is scored as very positive. This means that if the word "love" appears in a tweet about the 2023 presidential election, the Sentiment intensity analyser identifies it as a sentiment-related word.

3.5 Data visualization

Data visualization is the process of representing data in a visual format to facilitate easier understanding. To provide a comprehensive view of the factors considered when determining the type of tweets to collect, a graph that offers deeper insights into the data was generated. In this project, word clouds for visualization purposes were specifically utilized. A word cloud is an illustration of the most frequently appearing words in a text or dataset. It is widely employed in text analysis to quickly identify keywords and topics within a given text. Creating a word cloud typically involves measuring the frequency of each word in the text or dataset and sizing the words proportionally to visually represent their importance. Less frequently used words are typically displayed in smaller font sizes, whereas frequently used terms are typically printed in bigger font sizes. It is also possible to easily spot patterns and trends in the data thanks to the word arrangement [55].

The sentiment analysis of the #NigeriaDecides2023 using Twitter data provides valuable insights into the sentiments expressed by the Nigerian public during this crucial electoral event. By analyzing the sentiments shared on Twitter, a deeper understanding of the opinions, preferences, and concerns of the electorate were gained. The results and findings can inform decision-making processes, guide political strategies, and contribute to ongoing discussions on the role of social media in shaping political landscapes.

By illuminating voter sentiment and public opinion, this study demonstrated how sentiment analysis research could have a substantial impact on elections. Sentiment analysis utilizes Natural Language Processing techniques to assess various sources of data, such as social media posts, news stories, and other content, to determine the underlying sentiment or emotion. This analytical approach proves valuable for political parties and candidates as it allows them to gauge public sentiment towards their policies and themselves before an election. By examining social media posts and news stories within the context of sentiment analysis, patterns can be identified in how people respond to political events and speeches. This information becomes a useful tool for political campaigns to adapt their strategies and tactics accordingly. For example, if sentiment analysis reveals that voters disapprove of a particular policy proposal, a campaign may choose to modify or abandon that idea to garner more support. Additionally, sentiment analysis enables campaigns to identify key themes that resonate with voters and adjust their messaging to highlight those concerns.

In general, sentiment research offers valuable insights into public perceptions and the overall mood of the electorate, aiding political campaigns in making informed decisions and refining their approaches to better connect with voters. It is important to note, however, that not all eligible voters express their thoughts on Twitter, and there is variation in the usage of hashtags. Therefore, the findings of this study may not fully align with the actual election outcome. Furthermore, the tweets analyzed in this study were not geographically specific, making it challenging to determine if they solely originated from Nigerian accounts.

Nevertheless, by providing insights into the networks, impressions, and public sentiments surrounding the three prominent presidential candidates during the #NigeriaDecides2023 election, this study has contributed to our overall understanding of the election process.

5.1 Limitation to the Study

Generalization of Results: Not all Nigerian voters may agree with the results of sentiment analysis on Twitter data. Twitter users are a particular segment of the population, so they might not accurately reflect the opinions of the general public. Interpreting the sentiment analysis results in light of this restriction while avoiding over-generalization is difficult.

Linguistic and Cultural Nuances: The sentiment analysis may encounter challenges in accurately capturing sentiment due to the diverse linguistic and cultural expressions specific to Nigeria. Variations in dialects, local languages, and the use of slang or colloquialisms may introduce difficulties in sentiment classification and interpretation.

5.2 Future research

Several suggestions for further study and advancements can be made based on the results and findings of the sentiment analysis of the #NigeriaDecides2023 elections utilizing Twitter data:

Incorporate Data from Multiple Social Media Platforms: It would be advised that in future studies, the data collection be expanded beyond Twitter to incorporate information from Facebook, Instagram, and online discussion forums. This would give an improved understanding of the attitudes and viewpoints of the people across different online channels.

Investigate Emotion Analysis Techniques: Future studies should consider combining sentiment analysis with emotion analysis techniques. Insights into the intensity and nuance of public opinion would be gained as a result, allowing for a fuller understanding of the emotional reactions connected to the feelings voiced during the elections.