Roman alphabet is often adopted as a means of writing for input data and for their internal processing in the case of mechanical processing of Japanese language. And yet it is hoped that the final results of processing are written at least in Japanese syllabary (kana), which is more familiar to Japanese than Roman alphabet. Then a program which transliterates Roman alphabet into kana is required. First, specific features of Japanese strings written in Roman alphabet and writing rules of special symbols and strings which are not to be transliterated are collected. It is followed by an outline of the transliteration procedure based on the specific features and the writing rules. Then the directions for use of the actually constructed transliteration program RKHKN are given. An example of using the RKHKN and examples of the processing by the program are also presented. Finally, RKHKN is briefly compared with the KANA, which is also a transliteration program written by Takeya. RKHKN, written in COBOL, describes transliteration rules rather procedurally. KANA, on the other hand, has transliteration rules in the form of finite automaton, or transition network, and the procedure which drives the automaton. RKHKN is implemented on FACOM230-28 and also on FACOM230-38S.