Sentiment Analysis On Current Political Topics In Pakistan’s Twitter User Bases

doi:10.21203/rs.3.rs-2095172/v1

Download PDF

Research Article

Sentiment Analysis On Current Political Topics In Pakistan’s Twitter User Bases

https://doi.org/10.21203/rs.3.rs-2095172/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

A micro-blogging social media offers a way to the people around the world to show their sentiments about any political, social and cultural issue of the time. In this paper, sentimental analysis approach has been used to analyse the positive and negative comments of Pakistani people about some top trending political issues of Pakistan between Januarys to July 2018. The collected data was processed by using Python program with the help of SQL query and database. We have used graphs and tables to present the data, collected under five hashtags; which were top trending political topics of that time. Results were analysed and compared with the present situation and the results were then presented with outcomes after analysis. It is clear from the results and comparison that social media has a strong influence in the present era and can be used as a predictor of any political situation prevailing in the country.

SQL

API

PML(N)

Twitter, a micro blogging site, created in 2006 has now become a social platform to express the views on any social and political issue. Within a span of seven years, it has become one of the most visited website. More than 300 million active users post, comment and express them on Twitter.com site. Tweets and replies to tweets can be sent by cell phones, text and work region client or by posting on the site with maximum limit of 140 characters [1].

Sentiment Analysis is an approach used, in analyzing sentiments of twitter users about political or social topics. These sentiments can be positive, negative or sometimes neutral. Words and phrases used in tweets express the sentiments and emotions of people with particular mindsets. Machine Learning (ML) is a technique that uses artificial intelligence of machine to analyses and calculate data for future predictions. In business and other affairs, this is also known as prophetic research. The purpose of this research is to analyses the views and sentiments of twitter users on top trending political issues of Pakistan in a specific time frame. We have tried to relate the outcomes of our study with real life by keeping in mind the objectives and purpose of our research, related to the sentiments of twitter users about the political scenario of Pakistan. The tools and software used; are Python program to collect and extract tweets related to our topic. Twitter API and SQL Lite 3 were used for establishing dataset and filtering tweets as English and non-English. Excel and Chorus were used for analyzing and arranging the collected data.

We also faced some challenges during this process of data collection that hamper the work i.e. the extensive use of Roman and Urdu tweets and abbreviations like lol, hahaha, oh and many others.

After the analysis of our findings, it was clear that in future work, we would also consider the Roman and Urdu tweets for much accurate and broad study in this field.

The related work done in the field of sentiment analysis is not less; as there are various research work that investigates Slant Analysis (SA) [2]. Supposition investigation (SA) is generally a new term in this field. Twitter with its limit of 140 – character for tweets allow its users to be precise in their emotions expressed. The best outcomes got in the characterization of emotions utilizing managed learning methods, for example, Naive Bayes and Support Vector Machines; however, the manual coding vital for the regulated methodology is extravagant. There have been some non-directed work approaches referenced in and semi-managed procedures can be improved with further research[3],[4]. Specialists, with new qualities and evaluating strategies frequently contrast their outcomes and starting execution. It is important to make formal correlations between these outcomes that get from various grouping qualities and methods to choose the best attributes and the most proficient characterization systems for the applications[5].

Throughout the most recent decade, there has been a generous increment in the measure of work done in pg. 2 the field of Sentiment Analysis (SA). The work in the field of Sentiment Analysis (SA) has generally taken after the accessible information, both as far as the work done and the concentration region like Opinion Mining, Sentiment Investigation, Opinion Characterization and Micro-Blogging. [6] Moreover, it is essential to standardize the words stretched, for example, cooooooll, sooooooo or happyyyyyy replacing the letters that occur commonly in sequence with perhaps a couple of letters[7].

We conducted our study by gathering data from twitter using Twitter API. It is not a new idea but we have to separate English tweets from other languages tweets, it was not an easy task. Manual work was also done for this purpose. The methodologies used for analysis were unique with research. We collected our data for research by keeping in mind the purpose of our study.

Data collection is not an easy process as it might appear at initial stage. We used the keyword #hashtags that were further passed on Python using twitter API and credentials. The data extracted from twitter was of many languages but we focused on English tweets, gathered and transformed into dataset. To filter non-English tweets, we used SQL query to filter out all those tweets. After filtering tweets, we created a dataset to classify those gathered tweets into positive and negative. On the other hand, we ignored neutral tweets because they had no impact on analysis and results. Tweets were extracted by using the Python Program. It streams the data with the given hashtag. After applying the program, the twitter started the collection of tweets against the passed # tags. SQLite is an associated title in the collection of procedures that uses an autonomous transaction SQL information machine, without a server, zero configurations. SQLite is the most implemented information in the world with additional applications that we are about to tell, along with many other high profile [8]. It is the most implemented information, embedded level SQL without server method.

There is a trade-off between memory usage and speed, SQLite typically runs earlier the additional memory you provide it. However, its performance is typically quite smart, even in low memory environments. Betting on however, SQLite is quicker than direct file system. After passing the hashtag, data was streaming. When the tweets stop streaming the data was saved into database SQLite.

In the next phase, we started converting our collected data in tabular form by using SQL Lite 3 database. The tweets were in raw form on which we had to elaborate the other non-required fields such as follower count, friends count, status count and others.

Up to this point, we were not sure about negative and positive tweets, the datasets that would be utilized for preparing and analyzing the tweets were of English and non-English.

As earlier discussed, we are focusing on English tweets only but the twitter users from all over the world use different languages to express their sentiments and opinions. For this purpose, we need to filter other language tweets with SQL Query.

We collected the tweets from January – July 2018 using Twitter API about five top trending political and popular topics of Pakistan. The tweets were collected by passing the keyword / hashtag to the transmission API. A hashtag is a word that uniquely identifies a topic on Twitter. Following are the top trending topics on which tweeter user express their sentiments at the time of the study

#SheikhRasheed
#HanifAbbassi ephedrine
#Election2018
#Massive Rigging
#SherAaRaha Ha

We gathered tweets against each topic as per heading which depicts in below Fig. 3.1.

The results show which were just predictions before; after analysis those were outcomes to prove the sentiment analysis of twitter users about top trending political issues of Pakistan. After the data collection and pre-processing, each tweet was classified as positive or negative based on the polarity of words used. The polarity ranges from + 5 to -5 where higher #Sheikh Rasheed #Hanif Abbasi ephedrine #Election 2018 #Massive Rigging #Sher Aa Raha ha 0 5000 10000 15000 20000 25000 30000 35000 40000 pg. 3 value of positive was an indicative of very positive and highest value of negative was an indicative of very negative tweet. The value 0 indicated a neutral tweet, which was neither positive nor negative. The blue and red lines were used for the positive and negative tweets respectively. The height of the lines shows the strength of the tweet. The terms positive or negative are relative terms and their definition varies from hashtag to hashtag.

All the labelling of tweets are done by polarity score from chorus using simple rule i.e. we will first check the relevance of tweet to our definition and if a tweet falls into our definition then in the next phase we will use chorus for its scoring. We also validated this method for performing manual analysis of randomly picked tweets from each hashtag

4.1. #SheikhRasheed

The Fig. 4.1 constructed under the hashtag #SheikhRasheed which shows the sentiments of the twitter user about him. He even filed a serious alleged case against one of the strongest opponent of Pakistan Muslim League (Nawaz) PML (N), Hanif Abbasi. We found the mixed reaction about him in the tweets collection but finally it was observed that 60% tweets were with positive sentiments about him or we can say in his favor. This shows that people like him as a politician. Below figure shows the percentage where red color indicates negative sentiments while green shows in favor.

4.2. #HanifAbbasiEffidrine

In this #tag tweets show sentiments of public in his favor or against as a famous party worker of PML (N). Negative tweets were against him and his party and positive sentiments were in his favor. We can clearly see here that most of the people showed displeasure and did not like him for taking this quota. These sentiments can also be an indicative of people’s distrust on his party and him. He was barred from participating in the election and his party ticket was given to another PML (N) candidate. Since positive sentiment was an indication of people’s distrust on him and party too, so PML (N) candidate lost his seat as was predicted in the trends from our study. Below figure depicts public sentiments against this hashtag.

4.3. #Election2018

It is shown in graph through tweets, about hashtag #election2018, the sentiments of people regarding casting vote or not. People were generally not in the favour of vote caste because of massive rigging in previous election. Our graph just shows the sentiments of twitter users in favour or against about vote cast. It was obvious from the results that the ratio of casting vote dropped down up to 3% in this election as compared to previous election 2013.

4.4. #Massive Rigging

The sentiments in tweets were represented by our next graph about #Massive Rigging in general election 2018 where people expressed their views by positive tweets against rigging and negative tweets show the rigging. For this purpose manual analysis of tweets were also carried out. At the end of results, it was indicative through our graph, there will be rigging in the Election but FAFEN report was in the view that these elections were more transparent than those of 2013 elections were

4.5. #SherAaRahaha

People expressed their views in the favor of or against Pakistan Muslim League Nawaz, a political party of worth under the hashtag #SherAaRahaHai. Our graph for this hashtag indicates the sentiments of people about this topic. Positive tweets show that there will a win for PML (N) whereas negative tweets were against the party. After manual analysis and polarity score it was clear that there was not much difference between positive and negative sentiments but the results show that there were no more positive sentiments for #SherAaRahaHai.

The tweet collection graph Fig. 4.6 below sums up all sentiment graphs. Each label in the graph shows the label for hashtag. Y-axis shows the count of processed tweets. It is important to note that the definition of positivity has already been discussed in terms of each hashtag separately. Green color indicates positive sentiments while red color is for negative sentiments and for neutral sentiments black color is used.

Below Table 4.1 depicts the clear comparison of our predictions and results of the study, which were carried out to show the sentiments of people about election 2018 under five hashtags. It is very much clear that our predictions were near to results as people by using social media express their views and sentiments in favor of or against any political topic and define a way for the future political scenario of any country. Our research is a clear indicative of it.

Table 4.1

Predictions and Results
Predictions and Results
#tag	Our study sentiment graph prediction	Results 2018
#Sheikh Rasheed	Sheikh Rasheed will probably win	Won Election 2018
#HanifAbbasi Ephedrine	People shown displeasure towards his scandal and he/party will lose	His party candidate lost and convicted against this case
#Election2018	Voters Turn out ration may decrease by good margin	Turn out ratio is 53.62% which was decreased about 3% reported by FAFEN
#Massive Rigging	There is more chance of rigging in election 2018	Transparent as compared to 2013 General Election reported by FAFEN
#SherAa Raha ha	PML (N) will be the ruling party for another five years’ tenure	Against as predicted

Sentiment analysis has now become an important source in decision-making. People depend upon it for future predictions and efficient working; yet we cannot say it is more than enough to rely upon its judgments because opinions and perceptions of people vary with the circumstances. Overall, we can predict or make up our mind for future according to people’s sentiments either positive or negative.

In this paper, we have used the Sentiment Analysis (SA) approach to predict the outcomes of general elections of 2018 with reference of five #hashtags. Our findings show that social media like twitter helps to establish a sound perception of the social, political and cultural issues by analyzing the thoughts and feelings of people with respect of their comments. Our research shows that only 5.43% people commented in English language; rather people prefer to comment in their native languages or Roman Urdu and abbreviations. In future work, Roman Urdu and other languages can be also included for broader research and analysis. These types of studies and research open new vistas for researchers to further exploration in this field of sentiment analysis of all twitter users irrespective of any language usage. As the languages in tweets and retweets are more, the studies can be carried out in wider sense. This study is also a base for further work in sentiment analysis of social media users about any topic famous or notorious.

In addition, this research can make some sound grounds for further predictions by using social media like twitter.

ACKNOWLEDGMENT

I solemnly acknowledge that this research belongs to me.

AUTHORS’ NOTE

The authors declare that there is no conflict of interest regarding the publication of this article. Authors confirmed that the paper was free of plagiarism.

Twitter - Wikipedia
Liu, Bing, and Lei Zhang. ‘A survey of opinion mining and Sentiment Analysis (SA).’ Mining text data. Springer, Boston, MA, 2012. 415–463.
Turney, Peter D. ‘Thumbs up or thumbs down? Semantic orientation applied to unsupervised classification of reviews.’ Proceedings of the 40th annual meeting in association for computational linguistics. Association for Computational Linguistics, 2002.
Polk, Alexander, and Patrick Paroubek. ‘Twitter as a corpus for Sentiment Analysis (SA) and opinion mining.’ LREc. Vol. 10. No. 2010. 2010.
Diakopoulos, N. & Shamma, D., 2010. Characterizing Debate Performance via Aggregated Twitter Sentiment. Proceedings of the 28th int. conference on Human factors in computing systems. Go, A., Bhayani, R. & Huang, L., 2009. Twitter Sentiment Classification using Distant Supervision.
Larsen, Peder Olesen, and Markus von Ins. ‘The Rate of Growth in Scientific Publication and the Decline in Coverage Provided by Science Citation Index.’, Scientometrics 84.3 (2010): 575– 603. PMC. Web. 25 Sept. 2015.
Selmer, Øyvind, et al. ‘NTNU: Domain semiindependent short message sentiment classification.’ Second Joint Conference on Lexical and Computational Semantics (* SEM), Volume 2: Proceedings of the Seventh International Workshop on Semantic Evaluation (SemEval 2013). Vol. 2. 2013.
A. Giachanou and F. Crestani, ‘Like It or Not: A Survey of Twitter Sentiment Analysis Methods,’ ACM Comput. Surv., vol. 49, no. 2, pp. 1–41, 2016.
Shiels, Maggie. "Twitter co-founder Jack Dorsey rejoins company." BBC News 28.03 (2011): 2011.
Parr, Ben. "Twitter has 100 million monthly active users; 50% log in every day." Mashable, social media (2011).
Kurzweil, Ray, et al. The age of intelligent machines. Vol. 579. Cambridge, MA: MIT press, 1990.
Wilson, Theresa, Janyce Wiebe, and Paul Hoffmann. "Recognizing contextual polarity in phrase-level Sentiment Analysis (SA)." Proceedings of the conference on human language technology and empirical methods in natural language processing. Association for Computational Linguistics, 2005.
Turney, Peter D. "Thumbs up or thumbs down? Semantic orientation applied to unsupervised classification of reviews." Proceedings of the 40th annual meeting in association for computational linguistics. Association for Computational Linguistics, 2002.
Brody, Samuel, and Nicholas Diakopoulos. "Cooooooooooooooollllllllllllll!!!!!!!!!!!!!!: Using word lengthening to detect sentiment in microblogs." Proceedings of the conference on empirical methods in natural language processing. Association for Computational Linguistics, 2011.
Barbosa, Luciano, and Junlan Feng. "Robust sentiment detection on twitter from biased and noisy data." Proceedings of the 23rd international conference on computational linguistics: posters. Association for Computational Linguistics, 2010.
Agarwal, Apoorv, et al. "Sentiment Analysis (SA) of twitter data." Proceedings of the workshop on languages in social media. Association for Computational Linguistics, 2011.
Kouloumpis, Efthymios, Theresa Wilson, and Johanna D. Moore. "Twitter Sentiment Analysis (SA): The good the bad and the omg!." Icwsm 11.538-541 (2011): 164.
Go, Alec, Richa Bhayani, and Lei Huang. "Twitter sentiment classification using distant supervision." CS224N Project Report, Stanford 1.12 (2009).
Nagar, Anurag, and Hahsler Michael. "News Sentiment Analysis (SA) Using R to Predict Stock Market Trends." (2012).
Baccianella, Stefano, Andrea Esuli, and Fabrizio Sebastiani. "Sentiwordnet 3.0: an enhanced lexical resource for Sentiment Analysis (SA) and opinion mining." Lrec. Vol. 10. No. 2010. 2010.
Wilson, Theresa, Janyce Wiebe, and Paul Hoffmann. "Recognizing contextual polarity in phrase-level Sentiment Analysis (SA)." Proceedings of the conference on human language technology and empirical methods in natural language processing. Association for Computational Linguistics, 2005.
Barbosa, Luciano, and Junlan Feng. "Robust sentiment detection on twitter from biased and noisy data." Proceedings of the 23rd international conference on computational linguistics: posters. Association for Computational Linguistics, 2010.
Kouloumpis, Efthymios, Theresa Wilson, and Johanna D. Moore. "Twitter Sentiment Analysis (SA): The good the bad and the omg!." Icwsm 11.538-541 (2011): 164.
Barbosa, Luciano, and Junlan Feng. "Robust sentiment detection on twitter from biased and noisy data." Proceedings of the 23rd international conference on computational linguistics: posters. Association for Computational Linguistics, 2010.
Brody, Samuel, and Nicholas Diakopoulos. "Cooooooooooooooollllllllllllll!!!!!!!!!!!!!!: Using word lengthening to detect sentiment in microblogs." Proceedings of the conference on empirical methods in natural language processing. Association for Computational Linguistics, 2011.
Barbosa, Luciano, and Junlan Feng. "Robust sentiment detection on twitter from biased and noisy data." Proceedings of the 23rd international conference on computational linguistics: posters. Association for Computational Linguistics, 2010.
Howard, Philip N., And Aiden Duffy, Deen Freelon, Muzammil Hussain, Will Mari, and Marwa Mashhad. Opening Closed Regimes: What Was the Role of Social Media During the Arab Spring? Project on Information Technology and Political Islam Data Memo 2011.1. Seattle: University of Washington, 2011.
Girvan, M., Newman M.E.J. (2002). Community structure in social and biological networks, Proceedings of the National Academy of Sciences, 99 (12), pp. 7821–7826.
Yulong Pei, Nilanjan Chakraborty, and Katia Sycara. 2015. Nonnegative matrix tri-factorization with graph regularization for community detection in social networks. In Proceedings of the 24th International Conference on Artificial Intelligence (IJCAI’15), Qiang Yang and Michael Wooldridge (Eds.). AAAI Press 2083–2089.
J. Tang, X. Wang, and H. Liu. Integrating Social Media Data for Community Detection. Into Modeling and Mining Ubiquitous Social Media, 2012.
Mrinmaya Sachan, Danish Contractor, Tanveer A. Faruquie, and L. Venkata Subramaniam. 2012. Using content and interactions for discovering communities in social networks. In Proceedings of the 21st international conference on World Wide Web (WWW ’12). ACM, New York, NY, USA, 331–340.
Y. Ruan, D. Fuhry, and S. Parthasarathy. Efficient community detection in large networks using content and links. In WWW 13, 2013.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Sentiment Analysis On Current Political Topics In Pakistan’s Twitter User Bases

Status:

Version 1

Abstract

Figures

1. Introduction

2. LITERATURE REVIEW

3. Method

4. Results And Discussion

4.1. #SheikhRasheed

4.2. #HanifAbbasiEffidrine

4.3. #Election2018

4.4. #Massive Rigging

4.5. #SherAaRahaha

5. Conclusion And Future Work

Declarations

References

Additional Declarations

Status:

Version 1