Machine translation (MT) is always a challenging job. It is really difficult to build up a complete machine translation system for natural languages. Machine translation includes natural language understanding and generation. The proposed system represents a new solution for building a MT system for English to Bangla translation, by modifying the rule-based transfer approach of MT system. In machine translation the searching of word from the lexicon is a compulsory task, here this searching stage is utilized efficiently by proposing an intelligent integer based lexicon system, consists of a number of separate lexicons and an algorithm is also developed for searching words from the lexicon in order to accomplish the basic steps of machine translation.