LGRe: Latent Graph Recurrent Network for Document Ranking

Paper at DASFAA 2021

If you use this work, please cite as: bibtex

@InProceedings{LGRe,
author="Dong, Qian
and Niu, Shuzi",
title="Latent Graph Recurrent Network for Document Ranking",
booktitle="Database Systems for Advanced Applications",
year="2021",
publisher="Springer International Publishing",
pages="88--103"
}

Getting started

This code is tested on Python 3.6. Install dependencies using the following command:

pip install -r requirements.txt

We follow CEDR's data processing practices, and use Robust04 for example. After download this dataset, put it in data/robust04. Many of files for training and evaluation are available in data/robust.

qrels: a standard TREC-style query relevant file. Used for identifying relevant items for training pair generation and for validation(data/robust/qrels).

train_pairs: a tab-deliminted file containing pairs used for training. The training process will only use query-document pairs found in this file(data/robust/*.pairs). File format:

[query_id]	[doc_id]

valid_run: a standard TREC-style run file for re-ranking items for validation(data/robust/*.run).

datafiles: Files containing the text of queries and documents needed for training, validation, or testing. Should be in tab-delimited format as follows, where [type] is either query or doc, [id] is the identifer of the query or document (e.g., 132, clueweb12-0206wb-59-32292), and [text] is the textual content of the query or document (no tabs or newline characters, tokenization done by BertTokenizer).

[type]  [id]  [text]

Queries for WebTrack and Robust are available in data/robust/queries.tsv. Document text can be extracted from an index using extract_docs_from_index.py (be sure to use an index that has appropriate pre-processing). The script supports Indri indices. See instructions below for help installing pyndri.

Examples:

IndriBuildIndex data/robust-manifest.json
awk '{print $3}' data/robust/*.run | python extract_docs_from_index.py indri data/robust_idx > data/robust/documents.tsv

Training LGRe

To train LGRe, first download pre-trained Vanilla BERT and placed in data/cedr-models, then use the following command:

python train.py --fold n

Our experiment was conducted on two GTX1080Ti.

Acknowledgement

Some snippets of the codes are borrowed from NPRF, Capreolus, dl4marco-bert, SIGIR19-BERT-IR, CEDR.