Machine Learning Algorithm Identifies Tweets Sent Under the Influence of Alcohol

Interesting article posted recently in MIT Technology Reviews. What kind of metrics would help detect such tweets? Whether a picture or not is associated with the tweet Whether a link or not is associated with the tweet Number of typos for the tweet in question, compared with average for the user in question Frequency of tweets (sudden spike) for user in question Keywords (and mi-spelled keywords) typically found in such tweets across drunk users Replies / re-tweets from other twitters (volume, do they contain specific keywords?) Replies / re-tweets from other twitters (volume, do they contain specific keywords?) Which algorithm would you use?