Analyzing FIFA World Cup 2018 Tweets: Emoji Sentiment Analysis Project

This data science project analyzes approximately 530,000 tweets related to the #fifaworldcup2018 hashtag, focusing on emoji usage and sentiment. The project utilizes the Twitter API (or a Kaggle dataset) for data acquisition, followed by data cleaning and preprocessing. The core of the analysis involves extracting emojis using the 'emoji' module and performing sentiment analysis using the NLTK library. Sentiment scores are assigned to tweets, and these are then correlated with sentiment scores of emojis to gauge public opinion on the FIFA World Cup. The project includes feature engineering, the creation of a combined dataset, and data visualization using matplotlib to understand emoji popularity and sentiment trends. The final analysis aims to predict public sentiment and identify the most popular emojis associated with the event.

Emoji Data Science
Downloading and Reading file
In this document I’ll explain that how i have done twitter analysis of all the ( around 530000 )
tweets which was posted on the topic or with the topic #fifaworldcup2018 . For this analysis I
have used the twitter api to get all the tweets related to my point of concern ( i.e. tweets with the
#fifaworldcup2018 ), one another option of this analysis of the tweets is the dataset which is
available on the kaggle (with the title FIFA World cup 2018 Tweets, this file contains an
FIFA.csv file which is the required set of data which is needed ) , after I have downloaded the
dataset (or scrapped from the twitter using its api) the next task is data cleaning, one benefit of
using kaggle dataset is that cleaning of data is already done by the data provider. But for the sake
of efficiency I have checked and verified its cleaness by myself.
After all of these task I have moved to the next step ( i.e. the main steps all of the above steps are
the basics ) now I have to scrap all the emoji’s from the original tweet data , to do this task of
scraping the emoji’s from the text I have used the emoji module (this modules contains special
classes and methods which help us to scrap all the emoji’s from the string ) for my case I have
used a for loop and if statement to loop through each word of string ( tweets ) and using if
statement check that it is in emoji.UNICODE_EMOJI if it is in emoji.UNICODE_EMOJI it will
return that emoji . and finally we will get the emoji .

Now I have the original tweet and also I have the emoji , the next turn is to do the sentiment
analysis of the tweet and assign a polarity score and compare that polarity score with the emoji
Sentiment score and then assign an sentiment to my dataset . To do this all of the task I have
used nltk library which is designed for natural language processing . The main reason why I have
done sentiment analysis of the tweets by the people with #fifaworldcup2018 is because during
this analysis part my main aim was to predict the people view and thought on the fifa world cup
which was the most trending topic during the time when world cup is in progress . And also this
analysis will help to know the sentiments of people to my point of concern . And also this
analysis will definitely help me to predict that which emoji is the most people intrest ( i.e. the
emoji which is being used by most of the public , that is the most popular emoji which is most
people prefer or like to use most often ) , combing this emoji analysis with the sentiment analysis
of the tweets we can do better visualization for the data prediction, but before the visualization
part I have to do some feature engineering for our data . Feature engineering is the most
important factor the data science ( data visualization ) the more feature you add the more better
visualization can be done. The FIFA.csv file provides different columns related to the tweets but
for my point of concern the only important column is with the name original tweet all the
analysis of data visualization is done using only this column . using this column I have add 4
columns named as pos , neg , neu , compound this was done for the sentiment analysis from the
sentiment analysis ( sentiment analysis is done using nltk module for the natural language
processing ) the pos column contains the value of positive sentiment of the tweet ( positive
sentiment is between 0 and 1 ) , neg column contains the value of negative sentiment of the tweet
(negative sentiment is between 0 and 1 ) , neu column shows that how neutral the tweet is ( value
of neutral is also between 0 and 1 ) and the next comes the last column named compound this
shows the compound value of the tweet ( this is between -1 and 1 ) , i.e. pos shows the
percentage of positive words in a tweet and neg shows the percentage of negative words in the
same tweet and neu gives the percentage of neutral tweet and finally the last compound shows
the overall rating of the tweet it is between 0 and 1 . now it’s time to load emoji ranking dataset
The emoji ranking dataset is as shown below

This dataset contains 752 rows that is it contains 752 different emoji, the emoji is shown in the
column named Char despite this column the emojiranking dataset contain 4 more columns,
named as Neg , Neu , Pos , Sentiment score , this all the 4 column is made using sentiment
analysis by using nltk module, the four column has different meaning pos shows the percentage
of positive words in a tweet ( it is between 0 and 1 ) and neg shows the percentage of negative
words in the same tweet ( it is between 0 and 1 ) and neu gives the percentage of neutral tweet
( it is between 0 and 1 ) and finally the last compound shows the overall rating of the tweet it is
between 0 and 1 . now using all of this data present in the FIFA world cup dataset and the emoji
ranking dataset I have created an another dataset which contains the 6 column named as original
tweet , compound , neg , neu , pos , emoji , the dataset is as shown below :
This final dataset contains 530000 rows each with the different row ( that is the content of the
tweet in each row is different that is there is no any duplicate row in the this dataset ) and this
dataset the Orig_Tweet contains the original tweets of the peoples the pos column contains the
value of positive sentiment of the tweet ( positive sentiment is between 0 and 1 ) , neg column
contains the value of negative sentiment of the tweet (negative sentiment is between 0 and 1 ) ,
neu column shows that how neutral the tweet is ( value of neutral is also between 0 and 1 ) and
the next comes the last column named compound this shows the compound value of the tweet
( this is between -1 and 1 ) , i.e. pos shows the percentage of positive words in a tweet and neg
shows the percentage of negative words in the same tweet and neu gives the percentage of
neutral tweet and finally the last compound shows the overall rating of the tweet it is between 0
and 1 . generating this dataset took my device a long time ( over night ). So , it is very important
for me to save this generated dataset to my local drive , to save I have used pandas command and
saved it to my local drive so that when ever needed I can easily excess . I have saved this file
with the name fifasentimfile.csv and every time when I have to work I simply read this file from
my local drive using pandas , and saving the file to my local drive saved me a lot of time .

Visualizing emojis
Now its high time to visualize the emoji using ploting technique available in matplotlib ,
because visualization is the most important part of this project and I think that it is important part
of all the projects because a simple plot can explain a lots of thing that 100’s of lines of word
explanation can not explain . due to this reason I always tries to visualize most of the things for
proper explanation . and a plot can also be understood by an very less educated people and also
an uneducated people, so I have decided to plot the visualization using bar plot with the emoji
above it so that it will clearly show that how which emoji is used to which extent and also this
graphical visualization will explain that which is emoji is used most of the public that is which is
the most popular emoji . the tables in the data set is nice to look at but is there a more prettier
way to which we can visualize the data in the emoji , to do all of these task I have to first create
an grid and on that grid I will put all the things that I need ( I have decide to draw grid because
once a grid is created we can put all thing we need on that grid easily , that is we can easily add
and modify the grid and can get the desired plot ) , in this case I need to draw an emoji plot
which will show the number of emoji used per 1000 tweets of the data provided in the FIFA.csv
data . that is the y axis shows the count of the emoji and the x axis shows the type of emoji and
the whole plot shows that which emoji is used most in the analysis of the #fifaworldcup2018
tweets and this will help us to know the people sentiment using the sentiment analysis of the
emoji using emoji ranking . And finally my goal of the sentiment analysis of the tweets of the
#fifaworldcup2018 is completed.