An Advanced Quick-Answering System Intended for the e-Government Service in the Republic of Serbia

Abstract

Many of the services incorporated in the e-Government of the Republic of Serbia need a quick-answer system to meet the continually increasing demands of the citizens for easy, fast and effective obtaining of the requested information. However, the public administration of the Republic of Serbia contains a significant amount of unstructured data S. Nedeljkovic et al. An Advanced Quick-Answer System Intended for the e-Government Service of the Republic of Serbia-154-arranged in the documents. Thus, it is necessary to provide an automatic classification system based on the principle query-document. The question-answering (Q&A) system related to the Crime domain of the e-Government service of the Republic of Serbia represents a system for achieving the quick replies on citizens' questions. The Q&A system is based on the data mining, text mining, natural language processing, question answer, Bag of Words and N-gram analysis. A similarity measure (distance) is a significant parameter of the Q&A system due to its direct impact on searching speed and distance from wanted documents. Here, three most commonly used similarity measures are used: Cos, Jaccard and Euclid. The primary goal is to determine the similarity measure which provides the most precise results in the crime domain, and that similarity measure is used as a referent one. Due to the high importance of a similarity measure, we use the above three similarity measures, in the process of selecting the most appropriate similarity measure. The selection of the similarity measure is performed using the principles of redundancy and fault tolerance. Specifically, the principle of triple modular redundancy (TMR) with one voter is used. The proposed system is verified by the experiments with real citizen queries. The results show that the proposed system achieves good performance.

Enabling Public Security Text-Based Analytics: A Survey to Outline Research Directions

“…The performance improvement of the techniques used in text mining, in terms of metrics such as accuracy, precision, recall, and F1-score, is of great relevance for developing or improving frameworks and systems, appearing as a recurring expansion of objectives in proposals for future work. Analyzing techniques' performances allows researchers to incorporate and select different text mining approaches for a range of public security applications, making better decisions about which technique to use [25,86,[89][90][91][92][93].…”