A Filter-Map-Reduce Framework for Extracting Features From Log Files

• In Event: Thursday Roundtable Session 12:40 pm
  In Roundtable Session: Games for Cognition and Learning (Table 20)

Abstract

key challenge in designing educational games and assessments is extracting observable outcomes from the game logs. This paper provides a framework for extracting features consisting of three steps: (1) filtering—selecting relevant events,—(2) mapping—extracting key details from the messages,—and (3) reduction—combining the details into a single value. This framework is both optimized for using computer clusters for real-time scoring and is simple enough to allow designers to specify new features without needing to code.

Author

• Russell Almond, Florida State University