The website can be found here.
This simple web app takes either user-generated input text or direct quotes from the cartoon, SpongeBob SquarePants, and predicts which character among 10 possible characters was most likely to have said the quote. There are three options for a user to choose from:
-
Have the classifier predict the speaker of the input text you provide, and it will compare the result to the "true speaker" that you provide.
-
Have the classifier predict the speaker of a random test quote by a character that you specify.
-
Have the classifier predict the speaker of a random test quote by a random character.
Quantitative comparative analysis is done between the predicted speaker and the true speaker by comparying the weights of the n-grams present in the input text for each, as well as the most commonly spoken n-grams by each speaker.
If the application's guess is correct, then the comparative analysis will instead be between the true speaker and the speaker who the application believed was the 2nd most likely speaker of the input text.
Note that if the quote is user-generated, then the user also specifies which character said the quote. In this case, the user's chosen speaker is considered to be the "true speaker." Otherwise, a quote from the show is taken at random, and the true label of a test data quote is the true speaker.