Symbolic Prompt Tuning Completes the App Promotion Graph

Welcome to the official code repository for the ECML PKDD 2024 ADS paper "Symbolic Prompt Tuning Completes the App Promotion Graph". The data can be accessed through this link. Please download it and place it under the main folder.

The paper can be accessed through this link.

Code Structure

• main.py

  Running the model on the data stored in my_data folder.

• parser_hsf.py

  Setting the arguments for the running.

• prompts/

  Containing codes that generate prompts as needed.

  • gen_metarules.py generates rules according to metapaths for each query. The output file is all_pathcnt_2.pkl.
  • gen_prompts.py generates the related prompts. It can be set to metapath-only, embedding-based only, or both. This part relies on this repo for embedding-based prompts generation. To run it, please clone the above repo under the main folder, and set the paths accordingly. We have provided the pre-processed data in this link if the user would like to skip this step.

How to Run

Under the main directory, run python main.py --rand_format.