Roundtable: DELT-Evaluating the Effectiveness of Natural Language Processing Methods at Detecting Topics in Noisy Student Data

• In Event: Gallery Walk Showcase
  In Showcase Session: Gallery 07 - Approaches to Distance Learning Design and Engagement

Short Description

Unsupervised natural language processing methods have the potential of allowing automated detection of topics, sentiment, and other patterns in student-generated text, such as online discussions. However, little is known about how different methods are affected by the distinct semantic patterns that tend to be present in student-generated text. This study presents initial findings comparing the effectives of different unsupervised methods at detecting patterns on 300 simulated corpuses that mimic the features present in student-generated text.

Author

• Presenter: Tanner Matthew Phillips, Indiana University