Using certain algorithms to analyze in which time intervals messages are issued from a Twitter account, it is possible to differentiate whether the issuer is a person, a company or a computer program, regardless of the language used and the content of the tweet, according to a study from Imperial College London (UK) which publishes PLOS ONE magazine.
The researchers analyzed and processed 160,000 tweets sent by different types of users. And from the data obtained, they created algorithms to predict, with 83% accuracy, who was behind each account based on the time elapsed between two consecutive tweets. In addition, they have shown that by knowing when the last publication of a user of the microblogging platform occurred, it is possible to predict when a new message will be published .
The study reveals that, as a general rule, users of corporate accounts tweet more during work hours , while personal accounts are more active in the afternoon and at night. Accounts that use a computer program to launch their content usually post their tweets at regular intervals and constantly throughout the day, or have peaks of high activity at one or more times of the day. "This automatic classification of user types can be useful for research in the field of computer science, in advertising or even to carry out political campaigns", concludes Aldo Faisal, co-author of the work.
