Improving probabilistic models in text classification via active learning

HIGHLIGHTS

• What: To reduce labeling costs the authors propose a new algorithm for text classification that combines a probabilistic model with active learning. The authors demonstrate the performance of activeText in three ways. The authors show that this benefit is most pronounced when the corpus is unbalanced, and that upweighting keywords can boost further model performance. The second step involves actively selecting the documents that the model is most uncertain about and focusing manual labeling efforts among those documents (Hoi, Jin, and Lyu 2006).
• Who: MITCHELL BOSLEY and collaborators from the University of Canada have published . . .
  

   

  

  If you want to have access to all the content you need to log in!

  Thanks :)

  Username or Email

  Password

  Remember me

  Lost your password?

  If you don't have an account, you can create one here.