Traditional way of evaluating student translations has been criticized for being subjective and incomprehensive, which hinders rather than facilitates translation learning. This project proposes to incorporate corpus approach in translation evaluations. The aim of this study is to build an Error Annotated Learner Translation Corpus (EALTC), which consists of the translation work of about 150 translation students from the Department of Chinese, Translation and Linguistics. A corpus tool WordSmith 5.0 will be used to analyze the translation data. Both linguistic and non-linguistic information will be tagged to the errors including rendition errors (misunderstanding of source text, undertranslation, overtranslation and imprecise translation), linguistics errors (syntactic errors, semantic ambiguity, improper collocation, redundant words, unnecessary repetition, spelling, etc.) and miscellaneous errors. Based on this tagging process, errors will be categorized to provide the students with more detailed and tailored suggestions to help them reduce translation mistakes and develop effective translation strategies. The EALTC will serve as a useful tool in translation teaching and learning. Learners can use it to identify and analyze their own problems and learn from their peers. Teachers can illustrate translation theories, strategies and skills more clearly and convincingly with this pool of authentic examples.