Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Twitter Sentiment Analysis

155 views

Published on

Social Media micro-blogging networks such as Twitter is transforming the way Online Word-of-Mouth (WOM) is disseminated and consumed in the digital world.
In the present study, it is researched whether and how Twitter is influential as a WOM communication tool. In this study, Twitter social behaviour for the Hollywood movies is assessed across seven countries to validate the three basic blocks of the honeycomb model – sharing, conversation and reputation. Twitter behaviour and sentiments were studied for 30 movies in 22 different cities of seven countries with total tweets of 11.17 million tweets. To predict sentiments, a sentiment analyser using Naïve Bayes, MaxEnt and SVM machine learning techniques was developed. Prediction of sentiments was achieved with an accuracy of 90% for a Unigram language model in a 10-fold experiment.

Published in: Technology
  • DOWNLOAD FULL BOOKS INTO AVAILABLE FORMAT ......................................................................................................................... ......................................................................................................................... 1.DOWNLOAD FULL PDF EBOOK here { https://tinyurl.com/y8nn3gmc } ......................................................................................................................... 1.DOWNLOAD FULL EPUB Ebook here { https://tinyurl.com/y8nn3gmc } ......................................................................................................................... 1.DOWNLOAD FULL doc Ebook here { https://tinyurl.com/y8nn3gmc } ......................................................................................................................... 1.DOWNLOAD FULL PDF EBOOK here { https://tinyurl.com/y8nn3gmc } ......................................................................................................................... 1.DOWNLOAD FULL EPUB Ebook here { https://tinyurl.com/y8nn3gmc } ......................................................................................................................... 1.DOWNLOAD FULL doc Ebook here { https://tinyurl.com/y8nn3gmc } ......................................................................................................................... ......................................................................................................................... ......................................................................................................................... .............. Browse by Genre Available eBooks ......................................................................................................................... Art, Biography, Business, Chick Lit, Children's, Christian, Classics, Comics, Contemporary, Cookbooks, Crime, Ebooks, Fantasy, Fiction, Graphic Novels, Historical Fiction, History, Horror, Humor And Comedy, Manga, Memoir, Music, Mystery, Non Fiction, Paranormal, Philosophy, Poetry, Psychology, Religion, Romance, Science, Science Fiction, Self Help, Suspense, Spirituality, Sports, Thriller, Travel, Young Adult,
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here

Twitter Sentiment Analysis

  1. 1. Understanding Market Behavior in Microblogs (Twitter Sentiment Analysis) Umesh Hodeghatta, Ph.D March 5th, 2018 Email: umesh@mytechnospeak.com Phone: +1 408 757 0093 Web: http://www.mytechnospeak.com
  2. 2. Outline  Research Motivation ◦ Social Media Popularity ◦ How people express on social media  Research Objective  Analyzing Social Media Behaviour ◦ Data Collection ◦ Methodology ◦ Results and Discussion  Research Contribution  Conclusion  References
  3. 3. My Profile 3 20 + years of experience. Authored Two Books. Business Analytics Using R Introduction to Information Security AT &T Bell Labs, Cisco Systems, McAfee, Wipro MSEE (OSU) Ph.D (IIT Kharagpur) Specialize in Machine Learning and NLP Senior IEEE Member IEEE EXCOM ISACA Advisory Committee IT Advisor Government of India (Karnataka and Odisha) TechnicalAdvisory Member of the International Neural Network Society (INNS) India,Advisory member ofTask Force on Business Intelligence & Knowledge Management.
  4. 4. Social Media and Popularity  Social Media Wave ◦ Social media played an important part in Arab uprising, specifically in Egypt and Libya (Arnold, 2012, Ghannam, 2012) ◦ 2016 US election, campaign included the use of social media such as Facebook,Twitter ◦ South Africa,Taiwan,Tunisia,Turkey all saw the use of Social media to bring a major change in the political system (Tracey, 2012) ◦ Same is true in recent political election in India.  Supporters and other activists used the social media extensively to campaign 2014 election
  5. 5. People express opinions  Attempt to process (mine)Web ContentTo Actionable Knowledge for better decision making RealWorld Observed World Perceive Express Text Data (Social Media) Web Conten t Mining 1. Understand Language 2. Mood Of The observer 3. Infer about the real world 4. (Predictive Analytics) - predict certain interesting characters
  6. 6. Research Objective 1. Analysing Social Media (Twitter) Behavior in different countries 2. Sentiments on social media micorblogging site (Twitter) 3. Assessing sentiments across different geographical regions (countries) 4. Identifying social media impact on market
  7. 7. Movies sentiments on social media  The movies are of considerable interest to all the users community ◦ whether young or old or geographically at different locations  Large number of users interested in movies ◦ Substantial differences in the opinions expressed  Comparison of real world opinions vs actual box office collection  NOTE: Initial Pilot was tried on other products
  8. 8. ANALYSINGTWITTER (SOCIAL MEDIA) BEHAVIOUR Objective 1:
  9. 9. Analysing Twitter Behaviour Objective  Understand Twitter Behaviour. ◦ Conversation and Sharing proposed by Kietzmann et al. (2011).
  10. 10. Different Countries and Cities Country Cities Australia Sydney, Melbourne, Perth Canada Calgary,Vancouver,Totronto India New Delhi, Mumbai, Bangalre New Zealand Wellington,Auckland, Christchurch South Africa Durban, Cape Town, Johannesburg UK London, Glasgow and Birmingham USA San Francisco, Los Angeles, NewYork and Chicago
  11. 11. Hollywood Movies Genre Movies No. of Movies Action Fast and Furious, IFrankenstein, Jack Ryan: Shadow Recruit, Thor-The Dark World, White House Down, Wolverine, World War Z 7 Adventure Man of Steel, Iron Man 3, Percy Jackson Sea of Monster, The Hunger Games – Catching Fire 4 Animation Despicable Me 2, Free Birds, Frozen, Monster University, 4 Comedy Delivery Man, Last Vegas, The Hangover Part III, The Smurf2, We are the Millers,The Heat 6 Drama August – Osage County, Lee Daniel’s The Butler, The Fifth Estate 3 Sci-Fi Gravity, Elysium, Star Trek Into Darkness 3 Total 27
  12. 12. Twit Crawler
  13. 13. Twitter API (http://docs.twitter.com)  Following types of APIs ◦ Search API ◦ REST API and ◦ streaming API(http://docs.Twitter.com). Parameter Description q (required) A UTF-8, URL-encoded search query of 1,000 characters maximum lang (optional) This parameter restricts tweets to the given language – ‘en’ for English result_type (optional) This parameter specifies the type of results you would receive in the response. It supports three types:  recent – returns the most recent tweets  popular – most popular tweets  mixed – Includes both ‘recent’ and ‘popular’ types If you do not specify this parameter, search results returns ‘mixed’ type. count (optional) The number of tweets per page (maximum value of 100) Twitter Search API Parameters (http://dev.Twitter.com) Parameter Description geocode Returns tweets by users located within a given radius of the given latitude/longitude. The parameter value is specified by "latitude,longitude,radius", where radius units must be specified as either "mi" (miles) or "km" (kilometers). Example: 12.9715987,77.5945627,300mi (Bangalore) Twitter API Geo-Location Parameters
  14. 14. Twitter Output (open JSON file) { "iso_language_code": "en", "to_user_id": 0, "to_user_id_str": "0", "profile_image_url_https": "https://si0.twimg.com/profile_images/3118011529/0a61900f5a61811b776122e512bf3687_normal.jpeg", "from_user_id_str": "478733710", "source": "<a href="http://twitter.com/#!/download/ipad">Twitter for iPad</a>", "text": "Watching The life of Pi, absolutely stunning, but making me feel really sea sick!", "from_user_name": "Thatsnotmybookshop", "profile_image_url": "http://a0.twimg.com/profile_images/3118011529/0a61900f5a61811b776122e512bf3687_normal.jpeg", "id": 297471408617967617, "to_user": null, "entities": { "user_mentions": [], "hashtags": [], "urls": [] }, "to_user_name": null, "location": "Halifax, West Yorkshire", "from_user": "Emmasbookcorner", "from_user_id": 478733710, "metadata": { "result_type": "recent" }, "geo": null, "created_at": "Fri, 01 Feb 2013 22:28:05 +0000", "id_str": "297471408617967617" }, { "created_at": "Fri, 01 Feb 2013 22:25:47 +0000", verline wedo:pasadena United States:@scottjohnson; So, if Iron Man 3 came out a year or 2 ago, it would still be in theaters. Great movie. 1:29:21:205
  15. 15. Total tweets from Different Countries Movie Name USA (New York, San Francisco, Chicago, LA) Canada (Vancouver , Toronto, Calgary) UK (London Birmingham , Glasgow) India (Bangalore, Mumbai, Delhi) South Africa (Durban, Johannesbur g, Cape Town) Australia (Brisbane Melbourn Sydney) New Zealand (Auckland, Christchurc h Willington) Total Despicable Me 2 250002 134427 298712 16947 22093 59997 13134 838925 Elysium 53077 18092 110444 6523 11236 4559 2490 206421 Gravity 203301 125067 196375 90956 46576 66628 29935 758838 Iron Man 3 237894 33472 191097 7073 5949 4506 2602 482593 The Smurfs 2 23420 7830 33882 1734 1325 1497 665 70353 We Are The Millers 118438 37729 119462 6086 7195 3685 1965 294560 Wolverine 171018 43017 147212 15415 9054 22862 4892 413470 Delivery Man 56271 21288 67079 1957 1475 2850 1244 152164 Fast and Furious 6 74170 19899 174397 5964 15555 4092 2447 296524 Man of Steel 258235 89704 283015 36309 36367 31207 11300 746137 Star Trek into Darkness 100645 24931 137426 2350 1625 2869 1271 271117 The Fifth Estate 5236 5909 4151 267 19 1003 101 16686 World War Z 242052 96765 237785 10115 35190 20110 8139 650156 Lee Daniel's The Butler 27632 3984 15192 280 243 372 266 47969 Percy Jackson Sea of Monsters 2460 1106 3701 742 578 583 352 9522 The Hangover Part III 3734 1092 4634 322 344 217 28 10371 White House Down 172465 53136 189432 12210 25453 11181 6215 470092 The Heat 422416 275812 315204 118771 193257 154534 74431 1554425 Monsters University 226858 81405 241015 6959 5932 22105 6516 590790 August - Osage County 48102 14381 55447 1945 891 2073 1382 124221 Free Birds 41534 17490 34854 2354 1001 1801 627 99661 The hunger games : Catching Fire 117635 43594 112052 42703 6640 16201 6544 345369 Frozen 78633 24064 49878 4069 525 1892 952 160013 I, Frankenstein 61095 20941 49257 2237 576 1891 1316 137313 Thor - The Dark World 97423 29278 83882 7471 3322 4045 2058 227479 Last Vegas 140557 33119 80489 5008 2071 3139 2586 266969 Jack Ryan - Shadow Recruit 35239 7648 43206 1161 878 1685 721 90538 Grand Total (9.28 million tweets) 9289063
  16. 16. Twitter Behaviour 0 50000 100000 150000 200000 250000 300000 350000 400000 450000 DespicableMe2 Elysium Gravity IronMan3 TheSmurfs2 WeAreTheMillers Wolverine DeliveryMan FastandFurious6 ManofSteel StarTrekintoDarkness TheFifthEstate WorldWarZ LeeDaniel'sTheButler PercyJacksonSeaofMonsters TheHangoverPartIII WhiteHouseDown TheHeat MonstersUniversity August-OsageCounty FreeBirds Thehungergames:CatchingFire Frozen I,Frankenstein Thor-TheDarkWorld LastVegas JackRyan-ShadowRecruit USA (New York, Canada (Vancouver, UK (London India (Bangalore, Mumbai, Delhi) South Africa (Durban, Australia (Brisbane New Zealand (Auckland, Christchurch Willington)
  17. 17. Twitter Social Media Behaviour – Findings  Twitter users express their opinions on Hollywood movies  Twitter social media used across the world  Different countries’ users have different social media behaviour.  USA, Canada, and UK Twitter behaviour is different from other countries (more tweets than the other countries)
  18. 18. Managerial Implication  Promote product and CRM ◦ Companies can use social media to promote and get feedback on their products and services across different parts of the world and target products accordingly  Government Initiatives ◦ Governments can inform the public about their new policies, benefits of governmental programs to people, get buy-ins for any legislative changes  It is essential for companies to have a strong presence on social media to monitor their own products and services’ ‘conversation’ and what type of information is being ‘shared’
  19. 19. TWITTER SENTIMENTS ANALYSIS Objective 2:
  20. 20. Twitter Sentiment Analysis Objective:  Twitter Sentiment Analysis ◦ Validating reputation building block proposed by Kietzmann et al. (2011). Methodology • Machine Learning and NLP
  21. 21. Opinions  What other people think has always been an important piece of information for most of us during the decision-making process. Opinions are key influencers (Akcora et al., 2010)  Whenever consumer makes a decision, he/she often seek the opinions of others past experiences
  22. 22. Google product review
  23. 23. Positive or negative movie review?  Unbelievably disappointing  This is the greatest comedy ever filmed   It was pathetic.The worst part is about car race 23
  24. 24. What is Sentiment Analysis  Given a reviews, feedback, opinion the AI system should determines whether the review/feedback/opinion expresses sentiments of the reviewer ◦ Positive, Negative, etc.
  25. 25. INTRODUCTIONTO NLP 25
  26. 26. Natural Language Processing (NLP)  Linguistic analysis of naturally occurring texts for the purpose of achieving human-like language processing.  Computers to perform useful tasks involving human language ◦ Enabling human-machine communication ◦ Improving human-human communication 26
  27. 27. Goal of NLP 1. Paraphrase an input text 2. Answer questions about the contents of the text 3. Draw inferences from the text 4. Translate into another language 27
  28. 28. Sentiment Analysis - Approaches Two kinds of approaches: linguistic or nonlinguistic  Linguistic ◦ Syntactic and Semantic driven parsing  Syntax means ways that words can fit together to form higher level units such as phrases, clauses and sentences.  Nonlinguistic solutions based on statistics and machine learning techniques ◦ Language Modeling ◦ Naïve Bayes, SVM, etc.
  29. 29. NLP Difficult  Sentence might mean different things.  Ambiguity may be represented by different bracketing: ◦ Do you sell (Dell laptops) and (disk drives)? ◦ Do you sell (Dell (laptops and disk drives))?  As sentences get longer and grammars get more comprehensive 29
  30. 30. Statistical Language Model  Probability Distribution over word sequences ◦ Unigram Model ◦ Bigram Model ◦ N-gram Model  Generate text by generating each word independently P(w1 w2 w3 ….wn) = p(w1) p(w2) p(w3)….p(wn) {p(wi) } = p(w1) + p(w2) + p(w3)…p(WN) = 1 (N is size of vocabulary) Example: P(Leonardo DiCaprio is a brilliant actor) = p(“Leonardo DiCaprio”) + p(“is”) + p(“a”) + “(“brilliant”) + p(“(actor”) = 0.0002 * 0.003 * 0.0004 * 0.034 * 0.0043
  31. 31. Machine Learning Model – Classifier 31 Unlabelled Documents Labelled Documents ML Model Documents Labelled The goal is to build a model to automatically assign content categories to new unlabeled documents. ML Engine Predict New Data
  32. 32. Sentiment Analyser
  33. 33. Sentiment Analyser(Classifier) Sentiment Analyser Architecture Class Label Input Data Processing/ Cleaning Machine Learning Algorithm Prediction Model N-gram Input Data Processing/ Cleaning N-gram Class Label Training Prediction Raw Data
  34. 34. Data Cleaning and NLP  Stemming  Stop words  Synonyms (Tokenizing) • Removing user name • Removing timestamp • Removing URLs • Removing HASH • Removing double words • Removing white spaces • Removing punctuations • ……read chapter 6 3/26/2018 34 Robert Downey Jr.:Confidence:Don't forget to vote for RDJ, Iron Man 3 and @GwynethPaltrow in the #TCAs2013! :):2:4579:366:5534 "RT @50shadesofVini: Finished watching Life of Pi, what an emotional filmud83dude29"
  35. 35. Sample Sentiments Sentiments/ Statements Examples Positive Sentiments Django Unchained is legendary 10/10 ratings loved it "Just saw Les; this musical film literally took my breath away, beautifully filmed andamp; every actor/actress were just brilliant :)" Negative Sentiments django unchained and killing them softly sick movies "I have never been more stressed in my life watching a film than Argo and I know history well enough to know how it ends" life of pi may be the worst movie ive seen all year masturbatory waste of 2 hours that aimlessly bathes in its lack direction and content Cognitive Statements Django unchained makes record books as top US first day opening of all time for an r rated film "Interesting reading about the Iranian Revolution from the CIA's perspective #ARGO" Prospective Viewers Will be watching Skyfall • The words ‘loved’,‘beautiful’, represent positive sentiments, • while words like ‘killing’, ‘worst’, and ‘bad’ represent negative feelings. • Similarly cognitive statements contain information about the movie, for example, which theatre the movie is running in or the URL of the movie review. • Words like ‘watching’,‘watch’, represent prospective or potential viewers who have not seen the movie, or who want to watch the movie at a later date, or who are watching the movie right then, hence carrying no sentiments.
  36. 36. Sentiment Analysis of ‘Action’ Movies Countries Positive Negative Cognitive Prospectiv e Total Australia 27.68% 5.94% 3.91% 62.47% 100% Canada 69.76% 1.97% 19.30% 8.98% 100% India 33.72% 4.82% 57.89% 3.57% 100% New Zealand 80.34% 2.19% 14.79% 2.68% 100% South Africa 51.29% 9.42% 2.88% 36.41% 100% UK 54.84% 18.17% 24.81% 2.18% 100% USA 53.55% 15.09% 27.43% 3.93% 100% 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% Australia Canada India New Zealand South Africa UK USA TweetsinPercentage(%) Country Sentiment Analysis - Hollywood 'ACTION' Genre Positive Negative Cognitive Prospective
  37. 37. Sentiment Analysis of ‘Adventure’ Movies Countries Positive Negative Cognitive Prospective Total Australia 22.91% 9.48% 4.55% 63.06% 100% Canada 72.65% 1.94% 18.79% 6.62% 100% India 23.35% 15.50% 52.62% 8.52% 100% New Zealand 62.66% 2.99% 24.54% 9.81% 100% South Africa 47.33% 19.24% 3.38% 30.04% 100% UK 49.02% 18.30% 29.66% 3.02% 100% USA 33.40% 10.83% 51.53% 4.24% 100% 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% Australia Canada India New Zealand South Africa UK USA TweetsinPercentage(%) Country Sentiment Analysis - Hollywood 'Adventure' Genre Positive Negative Cognitive Prospective
  38. 38. Sentiment Analysis of ‘Animation’ Movies Countries Positive Negative Cognitive Prospective Total Australia 19.34% 1.31% 2.05% 77.30% 100% Canada 68.05% 0.55% 24.33% 7.06% 100% India 25.23% 4.18% 68.79% 1.79% 100% New Zealand 81.52% 0.75% 15.81% 1.92% 100% South Africa 52.26% 5.69% 2.02% 40.02% 100% UK 55.82% 8.70% 34.38% 1.09% 100% USA 51.03% 6.68% 39.39% 2.90% 100% 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% Australia Canada India New Zealand South Africa UK USA TweetsinPercentage(%) Country Sentiment Analysis - Hollywood 'Animation' Genre Positive Negative Cognitive Prospective
  39. 39. Sentiment Analysis of ‘Comedy’ Movies Countries Positive Negative Cognitive Prospective Total Australia 22.61% 2.53% 6.20% 68.66% 100% Canada 72.18% 1.14% 20.94% 5.74% 100% India 25.65% 1.68% 66.92% 5.75% 100% New Zealand 72.21% 1.31% 22.27% 4.21% 100% South Africa 61.05% 5.89% 0.78% 32.28% 100% UK 60.56% 11.36% 18.28% 9.80% 100% USA 57.39% 22.02% 16.85% 3.74% 100% 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% Australia Canada India New Zealand South Africa UK USA TweetsinPercentage(%) Country Sentiment Analysis - Hollywood 'Comedy' Genre Positive Negative Cognitive Prospective
  40. 40. Sentiment Analysis of ‘Drama’ Movies Countries Positive Negative Cognitive Prospective Total Australia 13.68% 2.49% 5.98% 77.85% 100% Canada 66.70% 1.18% 27.37% 4.74% 100% India 22.36% 0.46% 75.25% 1.93% 100% New Zealand 62.05% 2.03% 21.92% 14.01% 100% South Africa 54.51% 4.47% 13.40% 27.61% 100% UK 61.14% 9.84% 27.50% 1.52% 100% USA 39.44% 13.05% 43.75% 3.76% 100% 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% Australia Canada India New Zealand South Africa UK USA TweetsinPercentage(%) Country Sentiment Analysis - Hollywood 'Drama' Genre Positive Negative Cognitive Prospective
  41. 41. Sentiment Analysis of ‘Sci-Fi’ Movies Countries Positive Negative Cognitive Prospective Total Australia 22.5% 6.6% 4.4% 66.5% 100% Canada 66.9% 1.6% 20.8% 10.8% 100% India 34.1% 5.6% 51.6% 8.8% 100% New Zealand 66.0% 1.3% 24.7% 8.0% 100% South Africa 52.7% 10.9% 6.5% 29.9% 100% UK 49.1% 13.9% 33.8% 3.2% 100% USA 45.1% 12.6% 35.7% 6.5% 100% 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% Australia Canada India New Zealand South Africa UK USA TweetsinPercentage(%) Country Sentiment Analysis - Hollywood 'Sci-Fi' Genre Positive Negative Cognitive Prospective
  42. 42. Findings (contd..) Positive Negative Cognitive Prospective Action 53.03% 8.23% 21.57% 17.17% Adventure 44.47% 11.19% 26.44% 17.90% Animation 50.47% 3.98% 26.68% 18.87% Comedy 53.09% 6.56% 21.75% 18.60% Drama 45.70% 4.79% 30.74% 18.77% Sci-Fi 48.06% 7.49% 25.34% 19.11% Genre-wise Sentiment Analysis (all countries)
  43. 43. Findings (contd…) Positive Negative Cognitive Prospective Australia 21.45% 4.73% 4.51% 69.31% Canada 69.37% 1.39% 21.92% 7.33% India 27.39% 5.37% 62.18% 5.06% New Zealand 70.80% 1.76% 20.67% 6.77% South Africa 53.20% 9.27% 4.83% 32.71% UK 55.09% 13.37% 28.07% 3.47% USA 46.66% 13.38% 35.78% 4.18% Country-wise Sentiment Analysis (all genres)
  44. 44. Findings (Accuracy of Classifier) Australia (%) Canada (%) India (%) New Zealand (%) South Africa (%) UK (%) USA (%) Naïve Bayes 84 88 82 82 84 90 88 MaxEnt 44 46 42 44 46 46 50 SVM 42 42 42 42 42 42 42 82% - 87% (Dave et al., 2003; Pak and Paroubek, 2010; Bollen et al., 2011; Rui et al., 2013)
  45. 45. Measuring System Performance  Precision: % of selected items that are correct Recall: % of correct items that are selected GOLD System F 2 P. R P + R =
  46. 46. Output of Sentiment Analyser(Canada)
  47. 47. Output of Sentiment Analyser (USA)
  48. 48. Findings and Conclusion  Twitter sentiment vary from one country to another for the same Genre  Able to obtain higher accuracy by the combination of Statistical Language Modelling and NLP ◦ (82% - 87% by Dave et al., 2003; Pak and Paroubek, 2010; Bollen et al., 2011; Rui et al., 2013),  Classification of Tweets more than two classes possible
  49. 49. Managerial Implications  Knowing customers’ perceptions, opinions, and sentiments is becoming increasingly important in the current days of social media  Design or improve your products / services based on customer needs and desires  Plan your marketing strategy
  50. 50. RESEARCH CONTRIBUTIONS
  51. 51. Contribution (unique than previous studies)  Twitter Social Media follows Honeycomb Model – Sharing, Conversation and Reputation  Twitter Behaviour across different countries  Across 7 countries  Sentiment classification based on market relevance ◦ Across 7 countries  Twitter sentiments impact on Market Behaviour  http://www.frrole.com
  52. 52. Other Observations  UK Language ◦ Brits use whole sentences without even spelling errors  Indians uses more ‘Local’ language ◦ Bindas, Badmash, etc  USA uses more slangs than other countries  AustralianTweets are difficult to interpret
  53. 53. References  Storify compilation by Arnold, David. "Video News & Comment About Controversy & Future of Syrian’s Citizen Journalism." Multiple original stories published on Storify. June 2012,  Howard, Philip and Muzammil M. Hussain.. (2011), "The Role of Digital Media." Journal of Democracy. July 2011  Ghannam, Jeffrey. (2012), "Digital Media in the ArabWorld OneYear After the Revolutions“, Center for International Media Assistance, the National Endowment for Democracy.  LaurenTracey, (2013), Will social media influence election campaigning in South Africa? http://www.issafrica.org/iss-today/will-social-media-influence- election-campaigning-in-south-africa
  54. 54. References  Neelamegham, R., and Chintagunta, P. (1999),“A Bayesian Model to Forecast New Product Performance in Domestic and International Markets”, Marketing Science,Vol. 18, No. 2, pp. 115-136.  Pai, P.Y., andTsai, H.T. (2011),“HowVirtual Community Participation Influences Consumer Loyalty Intentions in Online Shopping Contexts:An Investigation of Mediating Factors”, Behavior and Information Technology,Vol. 30, pp. 603–615.  Pang, B., and Lee, L. (2008),“Opinion Mining and Sentiment Analysis”, Foundations andTrends in Information Retrieval,Vol. 2, No. 1-2, pp. 1-135.  Pang, B., Lee, L., andVaithyanathan, S (2002),“Thumbs up?: Sentiment Classification Using Machine Learning Techniques”, In Proceedings of the ACL-02 conference on Empirical methods in natural language processing,Vol. 10, pp. 79-86.  Rui, H., Liu,Y., and Whinston,A. (2013),“Whose andWhat Chatter Matters?The Effect of Tweets on Movie Sales”, Decision Support Systems,Vol. 55, No. 4, pp. 863-870.  Smith,A. N., Fischer, E., andYongjian, C. (2012),“How does brand-related user-generated content differ acrossYouTube, Facebook, andTwitter?, Journal of Interactive Marketing,Vol. 26, No. 2, pp. 102-113.  Sochay, S (1994),“Predicting the Performance of Motion Pictures”, Journal of Media Economics,Vol. 7, No. 4, pp. 1-20.  Turney. P. (2002),“Thumbs Up orThumbs Down? Semantic OrientationApplied to Unsupervised Classification of Reviews”, In Proceedings of the Association for Computational Lingusitics (ACL), pp. 417- 424.
  55. 55. Q & A ThankYou Email: umesh@mytechnospeak.com Phone: +1 408 757 0093 Web: http://www.mytechnospeak.com

×