Page - 331 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

Figure2. SmartMP3playerarchitectureandBlockdiagramof theexperiment fordatacollection. Theywere also informedofhow their datawill beusedandaprivacypolicywas added to the app. Participants were able to download and install the mobile application on their cellphones.Whenregistering toparticipate in theexperiment, theyareasked tofill out a concent formwithin the application.During 2weekperiod all the tweets that the participantsmadewouldbecapturedthoughtheapplicationandsubsequentlyprocessed, asFigure2(B) shows. The complete set of data captured and saved by the application when logging a tweet included thefollowingcontextualattributes:accelerometer(X,Y,Z);currentactiv- ities of the user selected from the categories:working, reading, drinking,watchingTV, eating breakfast, eating, dining, traveling, celebrating, exercising, doing homework, in class, listening tomusic,playingandcleaning thehome;5IBMToneAnalyseremotions associatedwith the posted Tweet; 5 emotions self-evaluated by the users; the song the userwascurrentlyplaying;date; latitudeandlongitude; translationof thepublished text; lightlevelandenergyandpleasantnesscalculatedwith thefuzzymodelproposed in[18]. For thisanalysis, thedataweusedwasthe latitude, longitude,emotional responseof ToneAnalyzer, self-perceived emotions anddate of the publicationof the tweet.A5x5 gridwasused to spacially stratify zones of interest covering the city ofLeo´n, as shown in Figure 3.We calculated howmany tweets wheremade inside each cell of the grid, doing a frequency analysis of the 5 emotions in eachquadrant of theRulermodel. The samegrid based representationwasused to visualise emotions encounteredby students at ITL. 5. PreliminaryResults The following sections present and discuss the results which were obtained from the experiments outlined in section4. InFigure 3(B) eachmarker, represents a tweetmade in that locationandhis frecuencyprocecedwithIBMisshownin3(A).Textevaluatedby user is shown in table2. Within each cell a frequency count of themajor emotionmade in each tweet, and themost frequent emotion shown in the center of each cell is shown inFigure 4Aand B. 5.1. Analysisof emotional frequency in thecity InFigures4,5and6,Cells inwhichanemotionisnotshown, indicate thatnoparticipant activity with the system occured. Cells where the equal sign is shown, indicate there J.J.ManriquezSantosetal. /ACrowdsourcingBasedComputationalMethodology 331