Abstract:
"In the ever-evolving landscape of education, the emergence of AI-generated text presents a
perplexing challenge: how do we distinguish between human and machine-generated content?
This issue is particularly pertinent in fields where authorship is crucial for credibility and
trustworthiness. ClearifyDetect aims to address this issue by providing both accurate detection
and understandable explanations for AI-generated text.
To address this challenge, ClearifyDetect explores advanced technologies and methodologies.
At its core lies a BiLSTM (Bidirectional Long Short-Term Memory) model, accurately trained
to discern patterns and nuances in text data. This model undergoes extensive training to
optimize its ability to accurately classify text as either human or AI-generated. Complementing
this capability is the Lime (Local Interpretable Model-agnostic Explanations) framework,
which enhances transparency and interpretability. Lime provides detailed explanations for each
prediction made by the BiLSTM model.
Through extensive training and validation processes, iterating 30 epochs, the BiLSTM model
achieves a strong accuracy in identifying AI-generated text. The incorporation of Lime's
explanations further enhances the system's utility by providing insights into the decision making process. As a result, Clarify Detect appears as a robust and essential tool for navigating
the complexities of AI-generated text, offering clarity and comprehension in an increasingly
complex landscape."
