Allie Wang – Analysis and Visualization, 50%
Adrian Stern – Data Mining and Processing, 50%
Final Project Deliverables:
Final Report: https://dl.dropbox.com/u/36595785/TwitterGotchi_Final_Report.pdf
We started the project with the goal of having a way to alert people if one of their friends on Twitter was showing signs of depression, or seemed to be making a call for help through their tweets. As we collected data, Hurricane Sandy hit the east coast. At this time we were streaming data and looking for Tweets that had hashtags related to sandy. By the end of Sandy we has a very large set of data related to sandy, because of this we shifted our focus to analyzing the impact of Hurricane Sandy through Tweets.
As our focus shifted so did our goals. Our goals were now the following: to prove that Hurricane Sandy affected people in the area hit by sandy, and did do in a negative way. As well as showing that this impact in their lives could be seen in their tweets, and could be proved scientific. If it could be proved we wanted to explore what problems and possible solutions could been seen through the tweets.
We meet our goals. We were able to see a difference in sentiment between the Sandy area and the Non Sandy area. We were then able to prove that the difference was statistically different using Chi-Squared test. We were also able to identify problems that people faced by using trending topics, we also could see coping methods using a similar technique. We further looked at this data by using key phrase detection to see what people were saying about problems, which gave further insight into the problems people faced.
Overall our focus changed and along with it our goals changed, but we were able to meet our new goals, and have a successful project.
Midterm Deliverable: https://dl.dropbox.com/u/36595785/AZW_AES_midterm.pdf
For TwitterGotchi, we try to identify users who exhibit depressive moods in their Tweets. The idea can be extended to the users’ social networks, which can be used to help therapists monitor patients’ depressive episodes; and in extreme cases, suicide prevention. In so doing, we draw from our research in Cognitive Behavioral Therapy. We call the project TwitterGochi after the Japanese toy, Tamagochi, in which the owner attempts to keep the Tamagochi happy and alive.
Pivoting off of our original idea, where we try to identify users who exhibit depressive moods in their Tweets, we attempt to classify Tweets as either negative or positive using machine learning classification models.
We will continue to call our project TwitterGotchi, for despite modifications, the project remains true to the sentiment behind our Twitter-based Tamagotchi.