Enhancement of stemming process for Malay illicit web content

Abstract

Web filtering system is one of the systems use to prevent users from can access any web pages that contain illicit contents. There are six (6) phases included in web filtering process. One of them is pre-processing phase. In this phase, there are three main activities included; HTML parsing, stemming, and stopping. The main focus in this research is stemming process. Stemming process is used to remove any affixes that attached together in the input words from web pages to produce the correct root words. To date, the existing stemming algorithm in Malay language; Othman’s stemming algorithm and Sembok’s stemming algorithm still produce errors in the result. Hence, the errors from both stemming algorithm were analyzed. Few features were created to encounter the problems occurred in existing stemming algorithm. There are initial checking with dictionary, implementation of Rule 2 and also checking with additional dictionary that contains the illicit words not included in the initial dictionary. These new features were added in enhanced stemming algorithm.In order to check the effectiveness of the new features added in the enhanced stemming algorithm, few tests were done to the sample of web pages. Based from the test, the result shows that only 11% corrected words produced if the test is done by without checking with initial dictionary and 72% corrected words produced if the process starts with initial checking with dictionary. The result for the test for implementation of Rule 2 shows that by using Sembok’s algorithm it produced only 17% corrected words compared with enhanced stemming algorithm produced 62% corrected words. As conclusion, the implementation of new features in enhanced stemming algorithm can reduce the errors produce in Sembok’s stemming algorithm.