A Study on the Application of Artificial Intelligence in e-Justice: Challenges and Opportunities

The legal sector, which was earlier under-electronic, has been completely transformed by artificial intelligence (AI) with natural language processing (NLP), and it is keen to embrace digital technology for greater performance. AI's possibility of decision prognosis for non-specialists, plaintiffs, and judges can help ease case congestion problems made worse by demographic expansion. Coding systems are essential for precise textual description and subsequent activities in e-justice, which is seen as a text classification/prediction issue. These systems function differently depending on the activity and database, capturing structure, semantics, and context. It isn't easy to choose the best algorithm, utilizing several assessment measures. Numerous prediction-oriented concepts were presented to address the E-justice challenge in India. It grows difficult to detect the changing character of crimes. E-judiciary is a strategy to lower crimes and prevent illegal activity. This paper introduces the stacking-oriented E-justice prediction technique (SEJPT), which uses SVM techniques and learning-based approaches to precisely estimate crime. The SVM method outperforms other AI models, including J48, SMO Naïve Bayes Bagging, and Random Forest, in terms of domain-specific setup. The ensemble algorithm with the largest coefficient of association, which has the least mean and absolute errors, functions better than the others in some circumstances. According to the study's statistics, the suggested technique's classification accuracy was 99.6%. It is discovered that the system has a greater prediction impact than the earlier studies that used E-justice datasets as norms. The findings also show that criminological concepts may be reconciled with any experimental statistics on crime. The suggested method was also successful in forecasting potential criminal activity. Additionally, it is suggested that the stacking ensemble system's estimation accuracy surpasses that of the particular classification algorithm.