We’re using the web service Topsy and a Ruby script to search for tweets that document police misconduct or benevolence, can be traced back to a specific officer, and are related to Occupy events. After cleaning the tweets of web noise, e.g. http://, we visualize the prominence of particular keywords associated with police misconduct. As a balancing counterpoint, we’re also searching for keywords associated with positive instances of police behavior.

Using IBM Research’s web service Many Eyes, Wordle, and  Tagxedo we’re also experimenting with methods to visualize recurring word patterns related to NYPD repression that viewers can interact with on the web.  Below we have featured some of our visualizations which were generated from Twitter mentions of police misconduct at Occupy. Click on them to access the interactive options.

Figure 1. Word cloud created from 39 tweets identifying an NYPD officer and documenting misconduct at Occupy Wall Street

Figure 2. Word cloud created from 23 tweets identifying an NYPD officer and documenting misconduct at Occupy Wall Street

And future-oriented, we’re also compiling a spreadsheet of badge numbers in anticipation of collaboration with the OWS Tech Ops working group. In doing so, we’ve noticed the importance of paying attention to the semantics of the tweets. How are the officers mentioned? What terms are used?

Figure 3. Word net created from 23 tweets identifying an NYPD officer and documenting misconduct at Occupy Wall Street

Figure 4. Stack hierarchy showing the percent of misconduct incidents by officer and type for the total collection

Figure 5. Matrix chart showing the frequency and types of misconduct by officer