5.2. Comparison with Variants of DAER To measure the effectiveness of our proposed DAER approach, our DAER was compared with four variants among three datasets, including DAER-C (i.e., no improvement and using as a blank control test), DAER-I (i.e., only adding improvements to the image feature extraction module), DAER-T (i.e., only adding improvements to the text feature extraction module), and DAER-I-T (i.e., adding improvements to both text and image feature extraction modules).
