The utterance units that serve as input to speech translation and/or spoken dialogue systems that handle spontaneous speech are not always sentences. However, the processing units of language translation are sentences. Since we do not have enough knowledge about the sentences of spoken languages, we use the term \language processing units" instead of \sentences." First, using conventionally interpreted dialogue data, we show that utterance units sometimes need to be divided into several language processing units, and sometimes need to be connected to make up a single language processing unit. Next, we propose a method of transforming utterance units into language processing units based on pause information and the N -gram of fine-grained part-of-speech subcategories. We confirm in experiments that our method yields good results. The technique reported in this paper has been introduced into the ATR-MATRIX speech-to-speech translation system from Japanese to English.