
A rule-based system predicts trademark/copyright case outcomes using word patterns, with moderate efficiency.
Authors
Somnath Mukhopadhyay, Department of Computer Science and Engineering, Assam University, Silchar, Assam, India
Jayati Mukherjee, Department of Computer Science and Engineering, Academy of Technology, West Bengal, India
Devangshu Das, Department of Computer Science and Engineering, Assam University, Silchar, Assam, India
Ashaawari Datta Chaudhuri, Lecturer, Jindal School of Banking and Finance, O.P. Jindal Global University, Sonipat, Haryana, India
Sunita Sarkar, Department of Computer Science and Engineering, Assam University, Silchar, Assam, India
Tamal Datta Chaudhuri, Bengal Economic Association, Kolkata, West Bengal, India
Kunal Paul, Department of Computer Science and Engineering, Assam University, Silchar, Assam, India
Summary
Legal cases involve specific terminology, use past judgments as references, and the entire legal process is expensive, both in terms of time and money. Further, it is not clear at the outset whether the expected judgment will prevail. In the context of trademark and copyright cases, the present paper develops a rule-based system that can be useful, for both lawyers and litigants, as an assisting tool to predict outcomes. The paper proposes a forecasting framework involving TF-IDF weighting scheme, Fuzzy C-means algorithm for clustering, the construction of decision trees using Gini Impurity Measure, and using Takagi–Sugeno fuzzy controller for efficient prediction. The dependent variable is binary, and we observe that the combination of specific words and their relative importance has a bearing on the judicial outcome. The paper goes beyond predicting outcomes based on relevant features, and suggests specific rules leading to outcomes of legal proceedings. Accuracy, Balanced Accuracy, Precision, Recall, and F-beta are used as forecasting efficiency metrics and the results indicate moderate forecasting efficiency.
Published in: Applied Soft Computing
To read the full article, please click here.