Detecting Illegal Wildlife Trade on Twitter

Deml, Gabriel

Student Work

Detecting Illegal Wildlife Trade on Twitter

Public

The Illegal Ivory Trade (IIT) is becoming an emerging concern in the world today. It is estimated that over 30,000 African elephants are killed by illegal trade each year. To help stop IIT, we investigated their existence and activities on Twitter. There have been attempts to do this, but they did not apply a strong machine learning model. Inspired by the belief that if poachers have no clients, they have no incentive to participate in IIT, we attempted to build frameworks that potentially remove the offending tweets. Insufficient data and crude algorithms limited prior works, thus not thoroughly exploring the possibility of an automatic deep learning framework for IIT detection. We, however, collected and annotated a sizable multimodal dataset (text, account profile, and image) of IIT tweets in this work. We then built a BERT-based deep learning model to identify these IIT tweets. Our model achieved an overall average accuracy of 94% and an average macro F1 score of 93% in a 10-fold cross-validation experiment setting on the dataset. We further provided some insights into the annotated data to shed more light on future works.

This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator