Our implementations for ADAPT, which has two key components: DUAL and ADEN.
Since the limited space in GitHub, we upload the demo data in google drive and BaiduYun (password: sad8) for running our code. Download the demo data, and put them into ADEN/Files/.
python 2.7, TensorFlow 1.12
DUAL is proposed to pre-train feature extractors of visual and textual features. The commands are as follows:
cd DUAL;
python main_dual.py
ADEN is proposed for storytelling recommendation by leveraging the cross-domain sequential behavior. The commands are as follows:
cd ADEN;
python main_aden_ft.py;
Note that python main_aden_ft.py uses the feature extractors pre-trained by DUAL model, but python main_aden.py is without using the pre-trained feature extractors. Our pre-trained models of feature extractors are in ADEN/Files/output/DUAL.