Audit System Development for Government Institution Documents Using Stream Deep Learning to Support Smart Governance

Abstract

Document audit system is a means of evaluating documents on the results of delivering information, administrative documentary evidence in the form of texts or others. Currently, these activities become easier with the presence of computer technology, smartphones, and the internet. One of the examples is the documents created by various government institutions whether local, city and central government. The instance is online-published documents that are shaded by certain government institutions. Before the documents are published or used as an archive or authentic evidence for reporting or auditing activities, the documents must go through the editing stage to correct if there are errors and deficiencies such as spelling errors or incomplete information. In the editing process, however, a person may not be able to escape from making mistakes that result in the existence of writing errors after the editing process before the submission. Word spelling mistakes can change the meaning of the conveyed knowledge and cause misunderstanding of information to the readers, especially for assessors or the audit team. Based on the problem, the researcher intends to assist the work of the audit preparation team in document analysis by proposing a system capable of detecting word spelling errors using the Dictionary Lookup method from Information Retrieval (IR) and Natural Language Processing (NLP) science combined with Stream Deep Learning algorithms. Dictionary Lookup method is considered effective in determining the spelling of words that are true or false based on Lexical Resource. In addition, String Matching method that has been developed can correct word-writing errors correctly and quickly.
Keywords: spelling mistake detection, dictionary lookup, audit of government institution documents, stream deep learning