AI. they dont hardcode specific words. Software learns from the examples given in the training phase then it builds its statistical model. It comes in many flavors, it easier to build than it sounds once you understand Statistics behind it. Just look up Sentiment Analysis.

posted 8 years ago in General Discussion

It is called supervised machine learning, sentiment analysis. Roughly it works like this: Software is initially trained with sample sentences labeled by people whether they are categorized toxic or not. Based on these training examples it builds vocabulary with toxicity "scores". When it is given a new sentence, by looking up the words in model that was previously built using training samples, it calculates the score of the comment.

posted 8 years ago in General Discussion

In some systems(unix linux etc) the date is measured in miliseconds starting from 01.01.1970 If the date of particular post is not available it just defaults to the 'beginning of the time' which is 1 January 1970 on the server moviechat runs.

posted 8 years ago in IMDb Information