ASR error correction is an effective method for optimizing ASR recognition results. The current mainstream ASR error correction system is mainly based on the encoder-decoder structure, by learning the mapping of incorrect text to correct text in a data-driven approach. Although this approach showed good results in ASR error correction, it mainly solves the spelling errors that occur at high frequencies, while struggling with low-frequency ASR entity errors, especially those unseen entities in the training set. Another downside of the end-to-end structure is that it may produce unreasonable correction results during decoding due to the lack of controllable mechanism. Therefore, how to design a controllable ASR error correction decoding strategy becomes a critical problem. In this paper, we introduce DDR-ECC—a Dictionary-Driven Entity Correction model with Controllable Decoding for Chinese ASR. We first retrieve relevant entities from an entity dictionary based on the similarity of pinyin and phonetic information, and then fuse them with ASR text for joint encoding. Next, a controlled decoding strategy is proposed to generate more reasonable decoding results. Experimental results on three publicly available datasets, Aishell, MAGICDATA \(_\text {SMALL}\) and MAGICDATA \(_\text {OOV}\) , demonstrate the effectiveness of our proposed method on ASR error correction, especially on ASR entity error correction. To be specific, compared to a solid encoder-decoder baseline(SC_BART) on the three datasets, the average SER and CER are decreased by 6.10% and 0.54% respectively, while the average F1 score of entity correction are improved by 9.06%. Finally, we define a new metric TD (tolerance degree) for measuring unreasonable decoding results, from which our proposed controllable decoding strategy can further improve the quality of error entity correction. Our code and data will be available at https://github.com/SeaEagleI/DDR-ECC .