A study on early decision making in objectionable web content classification

This study proposed early decision heuristics for objectionable content classification using an inverse chi-square classifier. The experimental results indicated that only examining the title plus 10% of a Web pagepsilas content can cost-effectively achieve an average precision of 93%. More importantly, the F₁ measure achieved its best when the title plus 60% of the body was examined. The proposed early decision making heuristics can serve as the trade-off baseline for real-time online filtering.