/HTPS

the model of HTPS on SIGIR2020

Primary LanguagePython

HTPS

Source code of SIGIR2020 Long Paper "Encoding History with Context-aware Representation Learning for Personalized Search". Due to two sub-models QDM and PLM have similar structures, we release the code of QDM model here which has a more classic hierarchical transformer structure.

Requirements

I test the code with the following packages. Other versions may also work, but I'm not sure.

  • Python 3.6
  • Pytorch 1.3.1 (with GPU support)

Data processing and train the model.

python dataset_new.py

Citations

If you use the code, please cite the following paper:

@inproceedings{ZhouDW20,
  author    = {Yujia Zhou and
               Zhicheng Dou and
               Ji{-}Rong Wen},
  title     = {Encoding History with Context-aware Representation Learning for Personalized
               Search},
  booktitle = {{SIGIR}},
  pages     = {1111--1120},
  publisher = {{ACM}},
  year      = {2020}
}