This paper describes a speech recognition research environment based on large-scale Japanese word and concept dictionaries. The dictionaries are currently under development at Japan Electronic Dictionary Research Institute (EDR). The research environment includes user-interactive capabilities, which enable efficient upgrade and revision of the dictionaries; and it has high-speed hardware, which supports real-time processing of speech input data. Within this environment, a combination of the two dictionaries will be applied to deal with the inherent ambiguities of speech recognition.