The gradual improvement and maturation of artificial intelligence technology has opened up new avenues for the application of corpora in translation teaching and learning in university English programs. This paper focuses on the intelligent detection of typical errors in student translations, leveraging the advantages of corpus-assisted translation instruction. For common omissions in actual English translation learning scenarios, we introduce information-based measures of sentence context variables and establish a set of synonyms for each sentence. Combining these with an XGBoost model, we construct an error detection model for omissions. Additionally, association rule mining is used to identify relationships between errors, generate an error node network, derive patterns among errors, and build an omission error detection model based on association rules. Using Chinese traditional culture as a case study, a Chinese-English bilingual corpus of Chinese traditional culture is established, and an AADAA teaching strategy based on the output-oriented approach is designed. Two classes of first-year English majors from K University were selected as control variables for the application experiment of the proposed model and teaching strategy. The post-test scores of the two classes in the translation course showed a significance level of less than 0.001, indicating a statistically significant difference, which validated the feasibility of the proposed model and teaching strategy in practical applications.
