Adult Content Detection in Search Engine Queries

Abstract

It is important to detect adult content in search engine queries in order to filter adult sites depending on the safe internet choices of the user. In this study, we investigate adult content classification in search engine entries for Turkish. Firstly, we collected and labeled data, and then we carried out classification experiments with both machine learning and deep learning methods. As a result of the experiments, we observed that deep learning methods performs better than machine learning methods. We obtained the best accuracy scores with the transformer based Electra model with 0.94 F1 score.