A Multi-Agent Based Approach to Short Message Service (SMS) Normalization System

Abstract

SMS language is characterized by fashion and user’s creativity but needs a transformation to proper English words or spelling to formulate natural language and text processing activities. The proposed application, Web Information Retrieval System Architecture Based on SMS (WIRSABoSMS), normalizes SMS language to retain its original syntactic structure. The concept of mobile agents in web technology was introduced as a medium to achieving Short Message Service (SMS) normalization. SMS normalization was carried out with the adoption of multi-agent technology, as agents are involved in character search, sort, and compare of the strings written in SMS form into its parental orthography. This architecture was designed based on web information retrieval system (IRS) in order to achieve SMS normalization. BLEU (bilingual evaluation understudy) was used to evaluate the quality of text. BLEU scores compare the human judgment with that of the machine translation using two set of corpora. The outcome of syntactic text message normalization recorded an average of 90% performance (for the corpus collected from researchers) when compared with the similar test conducted with human judgment using BLEU scores metric in an health-related domain.