This paper presents an automatic translation / speech synthesis system from written Esperanto to Japanese speech with a knowledge base editor incorporated to facilitate linguistic information implementation. The editor system enables the user to enter rules and to check their validity referring to the resultant synthesized speech. The system not only includes a novel efficient method of morpho-syntactic generation based on valency grammar, but also employs a simplified set of rules for morphophonological modifications. The three-tier rule system in phonology describes the rules classified according to the nature of the string on which they operate. The synthesis program converts the phonological surface into acoustic parameters, and then into speech.