Supported by the relevant theories of phraseology, this study examined the translation quality of three genres of explanatory, argumentative, and narrative essays at the phrase level and aimed to construct a translation quality assessment model. In this study, a total of six variables were extracted from both linguistic form and linguistic meaning in strict accordance with the phrase screening criteria, among which the linguistic form features contained a 2-4 gram match degree and the linguistic meaning features contained a part-of-speech tagged 2-4 gram match degree. The results showed that, first, bigram-related variables were the strongest predictors of translation scores for the three genres. The trigram-related variables were slightly weaker, and the fourgram-related variables were the lowest. The bigram match degree had the highest correlation coefficient of .752** with explanatory text translation scores. Second, all translation quality assessment models of the three genres fit well, with the highest correlation coefficient for the explanatory text model, R = 0.820, R2 = 0.673, followed by argumentative text, and the lowest for narrative text. This study realized the automatic assessment of the meaning and form of translations of different genres, which had certain theoretical and practical significance.