This portal performs authorship prediction using trained N-gram models across multiple dataset versions. Users may select predefined default models or customize dataset, model configuration, and feature sets for detailed evaluation. The system supports bulk document processing and enables parallel analysis across different experimental runs, allowing efficient comparison of results. Each submission undergoes structured feature extraction followed by model inference to classify content as AI-generated or human-authored.
The Default Models section provides quick access to commonly used dataset and model configurations that have already been predefined in the system. This option allows newer or more causal users to run authorship predictions immediately without manually selecting parameters.
The Customization section allows users to manually select the dataset version, trained model, and feature set used during prediction. This mode provides greater flexibility for experimentation and comparative analysis, enabling researchers to test different configurations and investigate how model choices affect prediction outcomes.