Abstract: Among all kinds of Chinese four-character idioms，the Parallel Four-Character Idiom（PFCI） is special and numerous.This paper introduces the research based on Conditional Random Fields（CRF） model which can retrieve PFCI from a POS-tagged corpus.The paper then analyzes the structural characteristics of PFCI and proposes an approach on recognizing PFCI in word-segmented corpora.By comparing its application on different corpora，the evaluation results show that this recognition approach maintains relatively high precision and good adaptability.