SNUH-mindspore

The MindSpore implementation of "Integrating Semantics and Neighborhood Information with Graph-Driven Generative Models for Document Retrieval" (ACL 2021).

The PyTorch implementation is available in this repo.

Dependencies

• Ubuntu-x86

• mindspore 1.2.0 CPU version

Datasets

We follow the setting of VDSH (Chaidaroon and Fang, 2017). Please download the data from here and move them into the ./data/ directory.

Run

The detailed running commands refer to run.sh.

Experimental Results

	16bits	32bits	64bits	128bits
Reuters	80.70	83.08	84.77	85.48
TMC	71.98	74.54	76.40	76.64
NG20	54.75	63.75	64.83	67.55

Cite

@article{zijing2021snuh,
  author    = {Zijing Ou and
               Qinliang Su and
               Jianxing Yu and
               Bang Liu and
               Jingwen Wang and
               Ruihui Zhao and
               Changyou Chen and
               Yefeng Zheng},
  title     = {Integrating Semantics and Neighborhood Information with Graph-Driven
               Generative Models for Document Retrieval},
  journal   = {arXiv preprint arXiv:2105.13066},
  year      = {2021},
}