Benchmark Using Social Media


Assigned Reading

  1. Introduction to sentiment analysis  Google► PubMed►  More►
  2. The Minute Survey  Read►
  3. Time to next complaint  Read►
  4. Sentiment analysis for satisfaction surveys  Read►


  1. How to calculate daily rates of complaints within Excel  Video► SWF► You Tube►
  2. Introduction to Minute Survey of satisfaction  Introduction► How it works?►
  3. Introduction to sentiment analysis You Tube► 

Narrated slides and videos require Flash.  Download►


Data:  To complete the project, you can collect comments made about a topic on the web.  In order to make the task easier, we have gathered from the web comments made about pediatricians and OB-GYN physicians in DC.  Download DC Pediatrics►  Download DC OBGYN►

Analysis: Two types of analysis are done.  Benchmarks are prepared by calculating time-to-next complaint.  See videos and papers assigned for this lecture to learn how you can do so.  Remember to set clinicians with no complaints to the time between oldest and most recent comments. 

The last task is to prepare a presentation about your work.  Your audience is directors of health administration programs across the nation.  Your presentation should clarify how sentiment analysis works and why this is a better method than alternative methods.   

Video:  If you are preparing the video, the audience are program managers and your task is to convince them of the value of sentiment analysis.  The video should have the following sections:

  1. Title slide:  Including your name is optional.   
  2. Student's voice:  Include a 15 second to 30 second video of a student complaining about a program.  You can find these on You Tube.  Download and edit what is on the web.  Use no personal videos.
  3. Description of sentiment analysis:  Describe the process of obtaining comments.  Give clear description of time to next complaint (with an accompanying visual) and use of automated text analysis (sentiment analysis) to identify what students are complaining about.     
  4. Sample Results:  Two sets of results are needed.  The first is benchmarked data.  Analyze the data for a specific programs  and provide a visual display of how that it stacks up against others. The second data refers to a summary of complaints.  Analyze up to 100 comments made for one of your programs and describe how these comments fit into various categories: 1) Instruction, (2) Course Content, (3)  Resources (4) Social life, (5) Academic rigor, (6) Other.  
  5. Discussion:  Provide brief discussion of why the approach is better than current program evaluations, e.g. by US World News.  Make sure that you do your homework regarding what is the current practice. 

The video should not exceed 10 minutes and should be posted on You Tube.  Please include the following statement in the description of the video:

"This video shows how minute survey and text classification programs (sentiment analysis) can detect service problems within an educational program.  This analysis relies on data taken from the web and is focused on health administration programs.  These data were analyzed as part of course on health information systems at George Mason University.  This course was taught by Farrokh Alemi, Ph.D., Phan Giang, Ph.D., and Hamid Jallal"

See example of student work in the context of analysis of sentiments towards physicians:  You Tube 1► You Tube 2►

This page was edited 05/16/2013 by Farrokh Alemi, Ph.D.  ęCopyright protected.