What We've Learned
A guided tour of interesting findings from the tweets we collected about US cities.
ChattyCity was designed and developed as a project to store, retrieve, visualize, and communicate data. Our project concept was to look at tweet sentiment (a general feeling or emotion, usually broken down as positive, negative, or neutral) based on tweeter's location (source) and where they are tweeting about (destination). We hypothesized that there is a bias for or against certain cities based on where the tweeter is from. For example, most residents in Seattle might have a positive sentiment about San Francisco, but residents of Los Angeles might generally have a negative sentiment toward San Francisco. Our project goals were to be able to answer questions from our data such as “which cities have positive sentiment toward San Francisco?” and to create an interface to allow a user to search our collection of tweets.
We recorded what 100 major US cities said about each other over a span of 3 weeks on Twitter, visualized the data, and created a search interface.
Twitter streaming API
July 7, 2014 through July 27, 2014
~15 million tweets (50GB of data)
Data Gathering and Cleansing: Python
Storage and Retrieval: AWS, Apache Solr
Data Visualization: Tableau, d3.js
For a more detailed description of the project, the tools used, and the source code, please visit our Github page.
Data Science graduate student at UC Berkeley and Astronomy Research Associate at UC Berkeley
Data Science graduate student at UC Berkeley and Market Intelligence Analyst at Autodesk
Data Science graduate student at UC Berkeley and Business Systems Analyst at Williams-Sonoma, Inc.
Image Credit: DaviDozz, Website Template: Agency by Start Bootstrap