California Institute of Technology (Caltech) and Stanford University researchers have demonstrated that machine learning algorithms can track evolving online social media conversations, which could eventually yield an automated method to detect trolling.
The technique is designed to overcome the ineffectiveness of current methods, which are either fully automated and non-interpretable, or reliant on a static series of keywords that can rapidly become obsolete.
The researchers employed a Global Vectors for Word Representation model, in which the distance between two words quantifies their linguistic or semantic resemblance, while also measuring relationships between keywords to determine context.
Said Caltech's Anima Anandkumar, "Hopefully, the tools we're developing now will help fight all kinds of harassment in the future."
From Caltech News
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA