zjunlp / rap

[SIGIR 2023] Schema-aware Reference as Prompt Improves Data-Efficient Knowledge Graph Construction

RAP

Code for the SIGIR2023 paper "Schema-aware Reference as Prompt Improves Data-Efficient Knowledge Graph Construction".

Requirements

Java 8 # for elasticsearch
elasticsearch==7.17.1

Retrieving for Reference

The reference store can be downloaded from here. Unzip the file and put the folder store/ into retrieval/ folder, and the final directory structure is as follows:

retrieval
├── store/
├── retrieve.py
└── retrieve_utils.py

For different base models, you can generate the reference by following codes:

cd retrieval/ 
python retrieve.py --base_model prgc

The parameter --base_model is for different base models, we can change it in prgc, relationprompt, t2e, degree.

For Text2Event and DEGREE, please follow the instruction README.md document in their corresponding folder to preprocess the datasets, and then generate the retrieved reference.

BaseModel

We plugged RAP to several base models, which can be seen in the folders below:

BaseModel
├── DEGREE
├── PRGC
├── RelationPrompt
└── Text2Event

The code of above base models are borrowed from their original codes with slight modifacations.

DEGREE : Please follow the instruction here.

PRGC : Please follow the instruction here.

RelationPrompt : Please follow the instruction here.

Text2Event : Please follow the instruction here.

rap's Issues

Document-level Relation Triple Extraction

Hello, can this model be used for document-level relation triple extraction? To fill the gap between sentence-level and document-level.

关于外部知识库EventWiki

您好，我看EventWiki原文并未开源数据，请问您是从哪里获取的？

retrieve.py

D:\Users\ShuaiqiDuan\Desktop\Research\DL\NLP_Pytorch\RAP\RAP\retrieval\retrieve.py:14: DeprecationWarning: Using positional arguments for APIs is deprecated and will be disabled in 8.0.0. Instead use only keyword arguments for all APIs. See elastic/elasticsearch-py#1698 for more information
if es.indices.exists(index_name):
E:\anaconda\envs\PyTorch\lib\site-packages\elasticsearch\connection\base.py:200: ElasticsearchWarning: Elasticsearch built-in security features are not enabled. Without authentication, your cluster could be accessible to anyone. See https://www.elastic.co/guide/en/elasticsearch/reference/7.17/security-minimal-setup.html to enable security.
warnings.warn(message, category=ElasticsearchWarning)
Traceback (most recent call last):
File "D:\Users\ShuaiqiDuan\Desktop\Research\DL\NLP_Pytorch\RAP\RAP\retrieval\retrieve.py", line 180, in
construct_store(es, data_name)
File "D:\Users\ShuaiqiDuan\Desktop\Research\DL\NLP_Pytorch\RAP\RAP\retrieval\retrieve.py", line 17, in construct_store
data = read_data_from_file(file_name=file_name)
File "D:\Users\ShuaiqiDuan\Desktop\Research\DL\NLP_Pytorch\RAP\RAP\retrieval\retrieve_utils.py", line 59, in read_data_from_file
inputs = open('store/' + file_name, 'r')
FileNotFoundError: [Errno 2] No such file or directory: 'store/nyt_store.json'
how can i solve it ?