Public Sentiments Analysis on Judicial Outcomes Against Expected Outcomes in Kenya

Abstract

The Kenyan judiciary has for a long time struggled to gain the confidence of the public. Judicial decisions in some cases do conflict with the expectations of the public and this serves to erode the public‟s confidence in the institution. The judiciary needs a means to gauge the public‟s opinions on ongoing cases in order to measure the deviation of its decisions from the expected outcomes from preliminary hearings on ongoing litigations. This should facilitate the judiciary to get in touch with the public‟s expectation and the effect of its decisions on the public. This should in the long run guide the judiciary in bettering its service delivery procedures to effectively serve the public. A means to mining of sentiments of the public to analyze them and understand their opinions is important in achieving this. A number of sentiment analysis algorithms exist. This research investigated the suitability of these algorithms by reviewing literature on their performance in similar problem domains i.e. text classification. The chosen algorithm was trained by help of Weka, starting with 70 initial of instances, of these instances 53% were correctly classified while 17 were incorrectly classified. This gave a 75% classification. Subsequent training of the model with the same number of instances gave 81% classification accuracy. This trained model was applied to public sentiments on three public cases. From the results, it‟s clear to note that in some cases public‟s opinions were not aligned with the judiciary‟s decisions, an indication of public‟s dissatisfaction in such decisions. In some cases, there was agreement. In one of the cases, the 2017 presidential petition, the percentage of those who approved the handling of the petition stood at 54% but that number drops to 34.5% after the judicial decision denoting disapproval of the decision by the public though they supported the process. The research used two models, a mathematical model of the algorithm used and a structural model to illustrate interaction of various components of the system prototype.