Tekin and Clark 1 Michael Tekin and Daniel Clark Dr. Schlitz Structures of English 5/13/13 Sarcasm in Social Media Introduction The research goals for this project were to figure out the different methodologies people use for conveying and understanding sarcasm over the Internet and social media sites. This research topic posed an interesting question. Sarcasm, being heavily conveyed in tone and facial expressions, should be difficult over the Internet, a faceless medium. The researchers looked into the annals of the Internet, books and microfilm to find both sarcasm and how people use it. On the microfilm (not listed as a source since the researchers make no reference to it) there was a short piece of literature about how women should not be doctors. The author said that if women became doctors then they should only take care of female patients. The two feminist male researchers were slightly outraged at the man from the 1970s who would say such sexist things. Then, realizing the heading of the book was Sarcasm and Satire, they realized they had fallen into the trap they had set out to learn more about. Disheartened at their silliness, they still continued on with renewed vigor, realizing that it is difficult to decipher sarcasm in print and they were on to something. Then the preliminary research methods involved the two researchers searching YouTube for videos on sarcasm. They found a clip from the popular television show Big Bang Theory, where Sheldon, an autistic-savant, was having extra trouble recognizing
Tekin and Clark 2 sarcasm. Penny, the beautiful blonde neighbor, barges into Sheldon and Leonard, the protagonist s, apartment angry about their antics from the night before. Penny says something sarcastic to Sheldon and Sheldon does not recognize it as sarcasm until Leonard holds up a sign that reads SARCASM. Our hypothesis was formed around that idea, that people need some sort of sign to recognize sarcasm. Big Bang Theory opened up a door for an interesting interpretation: Sheldon, the social media user, reads a message, post, or status (Penny) and cannot recognize the sarcasm if it were not for the sign (Leonard). This clip led the researchers on a road to discovering the different signs people used over social media. Thus, came the hypothesis that people consciously use several sarcastic signs to indicate that they are being sarcastic over the Internet. Background In Emoticons and Online Message Interpretation discussed that emoticons change the positivity or negativity of a sentence depending on the context of the sentence and the emoticon, showing the emoticon is an important part of digital culture. This study found that emoticons that are positive will make a message more positive and emoticons that are negative make a message more negative. It also claims that emoticons that contradict the meaning of the original message could be used to express sarcasm. Thanks for pointing that out. Making Sarcasm Accessible for All is an article that described sarcasm within deaf culture. In his study they used the subtitle sarcasm as a sign-post to indicate to the deaf audience the tone used was sarcastic. This article supports the idea that sarcasm can be indicated by such sign-post, and can be used to turn a phrase or sentence from serious to sarcastic. The methodology was based on the article How Sarcastic are You? Individual Differences and Verbal Irony. The question about
Tekin and Clark 3 posting a status as well as the idea of Twitter posts and the question of whether or not males are more adept at recognizing sarcasm was also came from this article. Last "Cruel to Be Kind and Kind to Be Cruel" states that knowledge of the person influence how people use and interpret sarcasm. Thus, since the methodology will involve an anonymous survey, this factor will have to be neglected from our results, even though it is a very important aspect of interpreting any statement. Methodology There were a total of 49 participants who took the survey. 21 were Bloomsburg University students who took a paper version of the survey and 28 participants that took it online. The paper survey was hand to a Structures of English class, and the online survey was advertised the online survey over Facebook via status updates and in Reddit under the subreddit section surveys. The online survey was conducted using survey monkey, and can be found at http://www.surveymonkey.com/s/ys7nsh7. Data was collected from surveys people took on paper, over Facebook and over Reddit. The surveys began with the basic questions about age, education, gender and race. They then moved on to general questions about what social media people use, how they use sarcasm, and how they recognize sarcasm coming from other people. The final question at the end of the free response section asked them to write down a sarcastic status, a short and descriptive statement, a term used online. Due to a limit in questions allowed by Survey Monkey, the third and fourth questions were combined as How do you recognize/convey sarcasm over these social media? There were also a total of 5 multiple-choice questions. In each question, one tweet was sarcastic and two were not. This section asks the participants to choose which twitter post was sarcastic and their
Tekin and Clark 4 answers were recorded. We did a chi-squared analysis to account for random guessing and used an Independent Two Tailed T-test to see if question 1 or two influenced how they answer (ex. gender) due to limited size on survey monkey the two of multiple choice question were excluded except for the chi squared analysis. All surveys accounted for all parts of the first question and second question were recorded based on their response. The third question (three and four on paper) was scored on what they answered. The categories they were scored under were: content, exaggeration, tone, jokes, emoticons/emojis, keyword/acronyms, follow up comments, grammar/structure, knowledge of the person, pictures/.gifs, and say so. These responses were scored under these categories, if their answer corresponds with a category. For example, if a participant had written I can tell a status is sarcastic by the poster s personality would be scored as knowledge of the person. Since participants often gave an answer that could be characterized under multiple categories, they were often scored multiple times. It was also noted if they did not know how to answer this question. These responses were scored as difficulty. The fourth question (fifth on paper) was analyzed using text analysis measuring the frequency of words used. Whether or not the participant also analyzed if they used the same variables they listed in the third and fourth question. The third and fourth question was then comparing how the respondents expressed sarcasm based on how they actually expressed sarcasm. This was accomplish by comparing how many responses that were scored for one feature in question four was also scored for the same feature in question three, and how many responses that were scored for one feature in question four was not scored for the feature in question. It was also noted if they had difficulty answer question three but still could answer question
Tekin and Clark 5 four. Last, the paper multiple choice question were analyzed to see if any of the features previously indentified in our study as being semantically important were present. Analysis A chi squared analysis was first used to determine whether the hashtag sarcasm was the only indicator of sarcasm used in the tweets used in the multiple choice question. The following chi squared values: The chi squared analysis for 5 multiple choice question produced the following values: Question # Chi squared value Question 1 42 Question 2 56.5 Question 3 5.4 Question 4 70.8 Question 5 47.7 Degrees of freedom 2 Fig 1. This is a chi-squared analysis testing if the high success rate was due to random guessing. All of the chi squared value observed in figure 1 produced a p values that were much lower than.01 for each question when using two degrees of freedom. Thus, the null hypothesis the following trends are due to random guessing. This indicates that the hash tag sarcasm was not the only variable used to indicate sarcasm. Next, to account for social-linguistic variables, an Independent Two Tailed T-Test on the variables indentified from question one. Due to insufficient sample sized we excluded the variables: ethnicity, age range, and education level. For example, 41vout of 49 participants identified themselves as being either white or Caucasian. The other 8 participants composed one of the following ethnicities: Ninja, Jewish, Hispanic/white, American, Portuguese, black, black/white and Irish. Thus, due to the low sample size of the Non-Caucasian, an accurate statistical analysis could not be performed. The only variable that could be used
Tekin and Clark 6 was gender. However, due to insufficient sample sizes, the one participant that did not know their gender, and the one participant that indentified as dragon were excluded. When comparing gender and average score the following results were found: Gender Average scores male 87.5% female 79.4% Total 81.6% Fig 2. The average score for every participant was 80.2%. The average sore in this figure, 81.6% was produced after excluding people who identified as dragon and did not know their gender. This table is comparing how and if gender effects your ability to recognize sarcasm based off the average success rate when answering the multiple choice question. When using Excel an independent Two Tailed T-Test was performed on males and females score, giving a p value of.445. This means that the following trend has a 44.5% chance of occurring due to random chance. Thus, there is no evidence to support that gender influence one s ability to recognize sarcasm. Moreover, due to lack of variation amongst other variables, the data collect can only accurately represent Caucasians between the age of 18-24 that was some form of college education since they are the most prominent group. On the other hand their isn t any evidence in this study to support the social-linguistic variable asked in question one affect one s ability to indentify sarcasm. Next, an independent Two Tailed T-Test was used to test if the social media the participant used affect their ability to recognize sarcasm. The social media sites the participants said they used were the following: facebook, instagram, pinterest, twitter, reddit, myspace and porn. First, porn was excluded since pornography is not a social media, and even if it was only 2 out of the 49 participant listed porn as a social media. Thus, this is not a sufficient sample size. Instagram, pinterest, and myspace were also exclude from this statistical analysis because of insufficient sample size. Facebook, on
Tekin and Clark 7 the other hand, was excluded because not enough participants didn t use facebook to do an accurate statistical analysis. However, people that used reddit, twitter, and tumblr had the following scores compared to people who did not use these sites. Uses twitter Success Yes 86.3% no 75.5% Total 80.2% Uses tumblr Success Yes 77.5% no 86.7% total 80.2% Uses reddit Success Yes 81.4% no 80% total 80.2% Fig 3. These tables express the overall average of each group of people who use and do not use each site and how it affected their ability to recognize sarcasm based on their success rate in the multiple choice questions. When an independent Two Tailed T-Test was performed on twitter user s and non-twitter user s scores a p value of.245 was calculated. When an independent Two Tailed T-Test was performed on reddit user s and non reddit user s scores, giving a p value of.905 was calculated. When an independent Two Tailed T-Test was performed on tumblr users and non-tumblr user s score, a p value of.337 was calculated. None of these p values fall below.05 so are not statistically significant. Thus, there is no evidence to support that using these sites provide any advantage of recognizing sarcasm over social media. Based off the free response the participants believed they expressed/recognized sarcasm using following ideas:
Tekin and Clark 8 pictures saying so just knew difficulty tone negative language stucture keywords emoticons irony humor exaggerqtion context 0 5 10 15 20 25 30 35 peranctage of responses scored under each feature Fig 4. The percentages are based off the percent of sarcastic statuses in question four that use each status. 7 statuses had to be exclude because they were unclassifiable. The fact that people said they used Structure, Keywords and Acronyms to express sarcasm over the internet indicates that people consciously add features to express sarcasm. This means that due to the written nature of language, expressing sarcasm over social media could be more performance based then competence based, like this study predicted. Either way, it shows that people are aware what makes a statement interpreted as sarcastic when used in social media. On the other hand, responses like tone, I read or I don t know indicates the recognition of sarcasm is also influenced by competence. To further examine whether or not people consciously recognized and expressed these features, the frequency of these variables was compiled in the following graph:
Tekin and Clark 9 Exaggeration Negative language Irony Context Saying so Structure Emoticons and emojis Keywords or acronyms 0.00 10.00 20.00 30.00 40.00 Fig. 5 The percentages are based off the percent of sarcastic statuses in question four that use each status. 7 statuses had to be exclude because they were unclassifiable. Tone and humor was excluded because there was no clear way to indentify these feature. Knowledge of the person was excluded because the survey was anonymous, and pictures were excluded because the participants were not given the ability to post a picture. Although these results alone do not indicate whether or not the participants were consciously using these variables, it is clear that all of these variables effect the semantic value of the sentence. When the frequency of each feature in question 4 was compared to how the participants claimed they recognized and displayed sarcasm in question we found the following results. Features Keywords or acronyms Features participants didn t say are used to express sarcasm but were used their sarcastic status Features participants claimed are used to express sarcasm and used in their sarcastic status 7 13 20 Total number of statuses that used these structures
Tekin and Clark 10 Emoticons and emojis 5 0 5 Structure 6 8 18 Saying so 1 2 3 Context 0 12 13 Irony 2 2 4 Negative language 1 2 4 Exaggeration 0 2 2 Fig.. Thus, surveys were only included in this if they had an identifiable response for questions 3 and 4. Therefore, column two and three do not always add up to column 4. The only feature that seemed to be only consciously used was emoticons. This indicates emoticons are probably a performance based aspect of language. On the other hand, context seems to be very competence based, even though 9 participants expressed its importance. While, the other features appears indentified in both question three and four appear to be a combination of both competence and performance. Moreover, 4 out of the 7 people who claimed they had difficulty recognizing/conveying sarcasm were able to post a sarcastic status. However, this could either be because the participants were not completely competent in sarcasm so mimic other people s method of sarcasm, or because they were not completely aware of how they were sarcastic but was competent enough to be sarcastic. Thus, due to the small sample sizes, and conflicting results, it is difficult to determine to what extend each feature is governed by competence or performance. However, as stated earlier, all of these features did effect the semantics of the sentence. Thus, the sarcastic statement in all of the 5 questions used in the paper survey to see what caused them to be interpreted sarcasm. All questions used keywords, while in one of the question there was one question that also used emoticons. This indicates, due to the total success rate of our participants 80.2%, keywords play an important role in recognition of
Tekin and Clark 11 sarcasm and the semantics of a sarcastic sentence. Lastly, a text analysis gave the following words as possible keywords possible Keyword Frequency Possible keywords frequency cont. Love 5 Great 3 Just 5 Much 2 Survey 5 Because 2 Don t 3 A lot 2 Really 3 Totally 2 Just kidding 3 Not 2 The status asked in question 4 was ran through a program called text stat, and the frequency of content morphemes was recorded and functional morphemes were excluded. These results suggest these words could be used to change a sentence that is semantically serious to one that is sarcastic. However, the sample size is too small to know for certain. Overall, the tests ran quite smoothly. There were few difficulties encountered. The difficulties encountered were not so much difficulties giant flying fire-breathing lizards, or the sub-category of our participants we will call dragons. There were some people who the researchers suspected did not take the survey seriously, including the person who named this category because he put his gender down as dragon. The researchers would have loved to find out whether dragons were good at understanding sarcasm, however the sample size (of one) was too small. So far, the results do not look too good (success rate: 0%). Conclusion Conveying sarcasm over social media does not seem to be as conscious of a decision as hypothesis predicted. Although are participants were fairly aware of how they express and recognize sarcasm. There were many features they used to convey and
Tekin and Clark 12 understand sarcasm they were not aware of. The main ones were keywords and acronyms. However, it was clear all the features we found contribute to the semantics of a sentence and if they are interpreted sarcastically. Lastly, although the sample does not compare enough variable in social-linguistic variables to say that these do not influence how someone s ability to recognize sarcasm, there is no evidence to say they do either. Plus, we are slightly concern on how some of our questions influenced one another. For example, did asking the respondents how they recognize social media influence their sarcastic satus. Thus, to further expand on any of these points were more specific and better advertised surveys are needed. For example, if we want to study how keyword influences someone s ability to be sarcastic, we could have to do another survey similar to the multiple choice question but these sarcastic statements that only a set amount keywords and contain none of the other features we indentified. If we wanted to see if this was conscious or not, we could do something similar to our sarcastic survey question, but we would need to ask what about their status makes it sarcastic. In conclusion, we found many interesting trends, that need further experimentation to confirm and refine to form any workable theories.
Tekin and Clark 13 Works Cited Derks, Daantje, Arjan E. R. Bos, and Jasper Von Grumbkow. "Emoticons and Online Message Interpretation." Social Science Computer Review 26.3 (2008): 379-88. Sage Journals. 10 Dec. 2007. Web. Fourney, David, and Deborah Fels. ""Thanks for Pointing That Out." Making Sarcasm Accessible For All." Proceedings of the Human Factors and Ergonomics Society (2008): 571-75. Web. Ivanko, Stacey L., Penny M. Pexman, and Kara M. Olineck. "How Sarcastic Are You? Individual Differences and Verbal Irony." Journal of Language and Social Psycology 23.3 (2004): 244-71. Sage Journals. Web. "sarcasm, n.". OED Online. March 2013. Oxford University Press. 13 May 2013 <http://www.oed.com/view/entry/170938?redirectedfrom=sarcasm>. Slugoski, Ben R., and William Turnbull. "Cruel to Be Kind and Kind to Be Cruel." Journal of Language and Social Psychology 7.2 (1988): 101-21. Sage Journals. Web.