A Two-Stage Ancient Chinese Poetry Correction Method

This paper focuses on the task of error correction for poems contained in long texts. The error correction of poems in long texts involves the recognition, positioning, and error correction of poetry texts. Compared with the error correction of single-sentence poems, it is more complicated regarding expression and grammatical errors. To solve this problem, we propose the Two-Stage Ancient Chinese Poetry Correction Method (TS-ACPC), which first identifies whether the news text contains poetry based on the pre-trained language model and then uses the precise matching method based on the vector space model to locate and correct the part containing ancient poetry in the text. In addition, the method of information retrieval is introduced to improve the speed of error correction of ancient poems in long text. In the final result, the accuracy of TS-ACPC’s ancient poem recognition is as high as 0.9964, and the accuracy of ancient poem error correction is 0.9275. The experimental results show that TS-ACPC is accurate and effective in correcting poems.