The bot finds positivity by looking at word usage, so right now its pretty simple. If your article has more words like 'yes', 'happy', 'good', etc., it will rate your article as positive. There are a lot of red herrings with this method - some posts have a lot of positive words but are not positive at all. Once I have enough feedback, I plan to cluster articles based on user feedback to improve positive article prediction.
You are viewing a single comment's thread from: