Communications of the ACM
Researchers Take Author Recognition to Neural Networks
Researchers at McGill University have developed a tool that can help differentiate Twitter users by their style of tweeting.
Credit: thestack.com
McGill University researchers say they have developed a forensic tool that uses neural network processing to help identify the subtle differences between authors' tweeting styles.
Their approach involves a set of parameters that also comprise the SMS text-messaging restrictions from which Twitter derived its original system.
A report examining the new technique proposed the use of neural networks to identify patterns that can be reinforced by the artificial restrictions.
The research abandons dataset-specific and manual processes in favor of automated analysis, and leverages the lexical n-gram model, which has been shown to be particularly effective in the identification of author gender, among other characteristics. The most useful identifiers include normal colloquialisms for everyday terms, and using Twitter-specific abbreviations and linguistic abuses that can prove characteristics of the user.
Although the current study is limited to English, the researchers acknowledge the approach will need to include other languages in order to achieve validity as a forensic tool.
From The Stack (UK)
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found