-JustDoIt-Nike-Tweets

Nike #JustDoIt Marketing Campaign Tweets Dataset: Project Overview (https://nrolle.github.io/-JustDoIt-Nike-Tweets/)

• Wrangled and tokenized the text of 5,000 tweets
• Visualized word counts using word clouds and bar graphs
• Performed sentiment analysis to identify the overall emotion of the text

This is a case study surrounding a marketing campaign conducted by Nike back in 2018. The dataset contains approximately 5,000 tweets in response to the marketing campaign. This particular marketing campaign is significant to me because it marked the day that Nike publicly stood with athlete and activist Colin Kaepernick in his efforts to bring attention to police brutality and the mistreatment of black and brown people in the United States. The objective of this project is to gain some insight in how the public reacted to Nike’s marketing campaign. I will be accomplishing this through tokenzing the text, sentiment analysis, and some graphical and numerical summaries of the dataset.

Code and Resources Used

RStudio Version: 1.4.1103 Packages: tidyverse, dplyr, tibble, tidytext, textdata, worldcloud2 Dataset(https://www.kaggle.com/eliasdabbas/5000-justdoit-tweets-dataset?select=justdoit_tweets_2018_09_07_2.csv)

Data Cleaning

• Selected only the columns deseriable for analysis and assigned them into a new data frame
• Removed tweets that did not relate to the topic (for example, renowned actor, Mark Hamill made a tweet using #JustDoIt that had nothing to do with Nike or the marketing campaign)

Removing Stop Words

• Stop words are common words that are not insightful for our analysis (such as the, like, and,etc.)
• Created a data frame composed of custom stop words
• A data frame of common stop words are available in the tidytext package as “stop_words”
• Combined the data frame “stop_words” and my “custom_stop_words”
• This is to ensure that R ignores words that are not insightful in our analysis

Tokenzing the dataset

• Some basic Natural Language Processing (NLP) Vocabulary
  • Every unique word is a term
  • Every occurrence of a term is a token
  • Tokenzing is the process of creating a “bag of words”
• Created a new object that gives us a column with a unique word in each row from each tweet

Counting the words

• Now that we imposed structure to the text we can count the words

Appending NRC Dictionary

• The Tidytext package includes 4 distinct sentiment dictionaries that works to tag each unique word with a sentiment
• I used the “NRC” dictionary to examine the overall emotional valence of the text
• The “NRC” dictionary defines each unique word as either negative, postiive, fear, anger, trust, sadness, disgust, anticipation, joy, or surprise

Sentiment Analysis

• By counting the number of occurreneces for each sentiment, we can get an understanding for the overall emotional valence of the text
• I created a bar graph to visualize what sentiment occurs the most

Top Twitter Handles Mentioned

• I created a data frame that told me what twitter handles were mentioned the most in these tweets
• I used a bar graph to visualize this

Trump, Kaepernick, Nike, Serena Williams

• In the previous bar graph we saw that the top twitter handles mentioned in this data set was Former President Donald Trump, Colin Kaepernick, Nike, and Serena Williams
• I thought it was be interesting to go further and look at the words and sentiment used when twitter users mentioned these influential figures

Trump - Word Cloud & Bar Graph

• I created a word cloud and bar graph to see the language used when twitter users mentioned @realDonaldTrump

Nike - Word Cloud & Bar Graph

• I created a word cloud and bar graph to see the language used when twitter users mentioned @Nike

Serena Williams - Word Cloud & Bar Graph

• I created a word cloud and bar graph to see the language used when twitter users mentioned @serenawilliams

Colin Kaepernick - Word Cloud & Bar Graph

• I created a word cloud and bar graph to see the lagnauge used when twitter users mentioned @Kaepernick7

Conclusion

• This marketing campaign was a successful brand marketing strategy for Nike
• At the time, not many companies had taken a stanceo n the topic of athlete activism in sport. For Nike to take a stance instead of remaining neutral was risky, however, it ultimately resonated well with its ocnsumes and reinforced the brand’s identity

#JustDoIt Case Study

Install necessary packages

install.packages(“tidyverse”) install.packages(“dplyr”) install.packages(“tibble”) install.packages(“tidytext”) install.packages(“textdata”) install.packages(“wordcloud2”)

Setting up the R Environment - Loading Libraries

library(tidyverse) library(dplyr) library(tibble) library(tidytext) library(textdata) library(topicmodels) library(tm) library(wordcloud2)

Read in File

JustDoIt = read.csv(“C:/Users/Nick/Desktop/justdoit_tweets_2018_09_07_2.csv”)

Data Cleaning

Tweets = JustDoIt %>% select(tweet_full_text, user_verified, tweet_retweet_count, tweet_favorite_count, user_screen_name, tweet_in_reply_to_screen_name, user_description, user_followers_count)

Tweets2 = Tweets[-c(1466),]

Number of rows in the Dataset

nrow(Tweets2)

Number of columns in the Dataset

ncol(Tweets2)

Overview of the Dataset

str(Tweets2)

Sample Tweet

sample = Tweets2 %>% select(user_screen_name, tweet_full_text)

sample_n(sample,1)

Removing words that are insigificant to the analysis

custom_stop_words = tribble(~word, “â”,”ðÿ”,”https”,”ï”,”t.co”,”realdonaldtrump”, “thinking”,”amp”,”kaepernick7”,”kaepernick”, “president”,”justdoit”,”nike”,”trump”, “people”,”ºðÿ”,”youâ”,”✔,”takeaknee”,”colinkaepernick”,”donâ”,”theyâ”, “ðÿš”,”potus”,”black”,”job”,”ve”,”shoes”,”white”,”unlike”,”money”,”real”,”weâ”, “wear”,”means”,”message”,”buy”,”stock”,”amendement”,”business”,”america”, “tweet”, “world”,”time”,”american”,”country”,”care”,”americans”,”ˆðÿ‘ÿðÿ”, “attention”,”bought”,”hey”,”canâ”,”ðÿ’ªðÿ”,”iâ”,”idk”,”paid”,”thatâ”, “wearing”,”commercial”, “ad”, “campaign”,”✚ðÿ”,”itâ”,”nikead”,”å”,”athletes”, “justdidit”,”marketing”,”pair”,”bogo”,”朔,”œðÿ”,”ä”,”15”,”ck”,”colin”,”æ”, “advertising”, “cqzvnmockn”,”doesnâ”,”customer”,”kapernick”,”nfl”,”æ­”,”apparel”,”ðÿ‘œ”, “ðÿ‘ÿ”,”nikeâ”,”nikecommercial”,”nikestore”,”serenawilliams”,”tonight’s”, “waiting”,”yâ”,”workers”,”weekend”,”taking”,”saturday”,”ðÿž”,”advances”,”7th”, “ceio2wcuyr”,”gkzrtyolqk”,”ll”,”lil”,”olc766xhvz”,”quicker”,”tonightâ”, “womenâ”,”themasb1”,”cutieðÿ’œ”,”ðÿ’ª”,”2”,”5i9egyh9tk”,”paylessinsider”, “payless”,”nflpa”,”nflcommish”,”gbmnyc”,”kstills”, “mosesbread72”,”1jedi_rey”,”havok_2o18”,”theswprincess”,”listentoezra”, “debbieinsideris”,”jeffbfish”,”knot4sharing”,”malcolmjenkins”,”kingjames”, “imwithkaep”, “matthewwolfff”,”kneel”,”jynerso_2017”,”jainaresists”,”b52malmet”, “deadpoolresists”,”debbiesideris”,”jynerso_2017”,”lady_star_gem”,”minervasbard”, “natcookresists”,”rebelscumpixie”,”sabineresists”,”trinityresists”,”drawing”, “batmanresist”,”brandontxneely”,”blue”,”captainslog2o18”,”earl_thomas”,”exercise”, “nateboyer37”,”plays”,”realtomsongs”,”tdlockett12”,”trisresists”,”xtxoan4y7d”, “ybbkaren”,”zmndpufdoh”, “president”,”police”,”football”,”gear”,”lord”,”god”, “military”,”catch”,”school”,”labor”,”wait”,”feeling”,”vote”,”pay”,”deal”,”words”,”coming”, “finally”,”shot”,”chance”,”guess”,”leave”,”change”,”mouth”,”fisa”,”1st”,”1”, “10”,”nikes”,”blah”,”orange”,”buying”,”single”,”anonymous”,”office”,”stuff”, “life”,”kids”,”lie”,”bad”,”btw”,”concept”,”level”,”players”,”kap”, “national”,”serve”,”social”,”sell”,”company”,”true”,”decision”,”watching”,”running”, “guy”,”ooh”,”f45”, “gum”, “chose”,”yr”,”uk”,”ðÿž”,”ev”,”front”,”head”,”chaserâ”, “dawn”,”jedimasterdre”,”land”)

stop_words2 = stop_words %>% bind_rows(custom_stop_words)

Tokenzing the Dataset

tidy_Tweets = Tweets2 %>% unnest_tokens(word,tweet_full_text) %>% anti_join(stop_words2)

Counting words

count_Tweets = tidy_Tweets %>% count(word) %>% arrange(desc(n))

Appending NRC Dictionary

sentiment_tweets = tidy_Tweets %>% inner_join(get_sentiments(“nrc”))

Counting sentiment

sentiment_count = sentiment_tweets %>% count(sentiment) %>% arrange(desc(n)) %>% mutate(sentiment2 = fct_reorder(sentiment,n))

Visualizing sentiment

ggplot(sentiment_count, aes(x=sentiment2, y=n, fill = sentiment))+ geom_col(show.legend = FALSE)+ coord_flip()+ labs(title = “Sentiment Counts Using NRC”, x= “Sentiment”, y= “Counts”)

Tibble to show each sentiment

word_counts = sentiment_tweets %>% count(word, sentiment) %>% group_by(sentiment) %>% top_n(10, n) %>% ungroup() %>% mutate(word2 = fct_reorder(word, n))

Visualizing each sentiment

ggplot(word_counts, aes(x=word2, y=n,fill=sentiment))+ geom_col(show.legend = FALSE)+ facet_wrap(~sentiment, scales=”free”)+ coord_flip()+ labs(title = “Sentiment Word Counts”, x=”Words”)

Top Twitter Handles mentioned

Replies = Tweets2 %>% select(tweet_in_reply_to_screen_name) %>% count(tweet_in_reply_to_screen_name) %>% arrange(desc(tweet_in_reply_to_screen_name)) %>% top_n(10)

Removing the Blank Row in Replies

Replies_2 = Replies[-c(12),]

Fct_reorder - Cleaning up for the Graph

Replies_3 = Replies_2 %>% mutate(n2 = fct_reorder(tweet_in_reply_to_screen_name, n))

Visualization of Replies_2

ggplot(data=Replies_3, aes(x=n2, y=n, fill = n2))+ geom_bar(stat=”identity”, show.legend = FALSE)+ coord_flip()+ ggtitle(“Top Twitter Handles Mentioned”)+ xlab(“Twitter Handle”)+ ylab(“Number of Mentions”)

Filtering for tweets that mention Donald Trump

Trump = Tweets2 %>% filter(Tweets2$tweet_in_reply_to_screen_name == “realDonaldTrump”)

Tokenzing text

tidy_Trump = Trump %>% unnest_tokens(word, tweet_full_text) %>% anti_join(stop_words2)

Counting words in the text

count_Trump = tidy_Trump %>% count(word) %>% arrange(desc(n)) %>% top_n(50)

Word Cloud for words used when mentioning Donald Trump

wordcloud2(count_Trump, rotateRatio = 0)

Reordering the Count Dataframe for bar graph

count_Trump2 = count_Trump %>% mutate(word2 = fct_reorder(word, n)) %>% top_n(50)

Visualizing the Count Dataframe

ggplot(count_Trump2, aes(x=word2, y=n, fill=word))+ geom_col(show.legend = FALSE)+ coord_flip()+ labs(title = “Words Frequently Used When Mentioning Trump”, x = “Frequency”, y = “Words”)

Filtering tweets that mention @Nike

Nike = Tweets2 %>% filter(Tweets2$tweet_in_reply_to_screen_name == “Nike”)

Tokenzing the text

tidy_Nike = Nike %>% unnest_tokens(word, tweet_full_text) %>% anti_join(stop_words2)

Counting the text

count_Nike = tidy_Nike %>% count(word) %>% arrange(desc(n)) %>% top_n(50)

Word cloud of the word count

wordcloud2(count_Nike, rotateRatio = 0)

Factor reordering the counnt data frame for visualization purposes

plot_Nike = count_Nike %>% top_n(10) %>% mutate(word3 = fct_reorder(word,n))

Visualization of word count

ggplot(data=plot_Nike, aes(x=word3, y=n, fill = word))+ geom_col(show.legend = FALSE)+ coord_flip()+ labs(title = “Words Frequently Used when Twitter mentioned @Nike”, x = “Words”, y = “Frequency”)

Tweets at Serena Williams

Serena = Tweets2 %>% filter(Tweets2$tweet_in_reply_to_screen_name == “serenawilliams”)

Tokenzing text - Serena Williams

tidy_Serena = Serena %>% unnest_tokens(word, tweet_full_text) %>% anti_join(stop_words2)

Counting tokenized text - Serena Williams

count_Serena = tidy_Serena %>% count(word) %>% arrange(desc(n))

Visualizing Text - Serena Williams

wordcloud2(count_Serena, rotateRatio = 0)

Factor reorder for word count data frame

plot_Serena = count_Serena %>% mutate(word4 = fct_reorder(word, n))

Visualization

ggplot(data=plot_Serena, aes(x=word4,y=n, fill=word4))+ geom_col(show.legend = FALSE)+ coord_flip()+ labs(title = “Words Frequently Used When Twitter Mentioned @serenawilliams in a Tweet”, x=”Words”,y=”Frequency”)

Filtering for tweets that mention @Kaepernick7

Kaep = Tweets2 %>% filter(Tweets2$tweet_in_reply_to_screen_name == “Kaepernick7”)

Tokenzing the text

tidy_Kaep = Kaep %>% unnest_tokens(word, tweet_full_text) %>% anti_join(stop_words2)

Counting the words

count_Kaep = tidy_Kaep %>% count(word) %>% arrange(desc(n)) %>% top_n(25)

Word Cloud

wordcloud2(count_Kaep, rotateRatio = 0)

Bar graph

plot_Kaep=count_Kaep %>% mutate(word5 = fct_reorder(word,n))

Visualization

ggplot(data=plot_Kaep, aes(word5, n, fill=word5))+ geom_col(show.legend = FALSE)+ coord_flip()+ labs(title = “Words Frequently Used When Twitter Mentioned @Kaepernick7 in a Tweet”, x=”Words”, y=”Frequency”)