PURE: End-to-End Relation Extraction

Related tags

Deep LearningPURE
Overview

PURE: End-to-End Relation Extraction

This repository contains (PyTorch) code and pre-trained models for PURE (the Princeton University Relation Extraction system), described by the paper: A Frustratingly Easy Approach for Entity and Relation Extraction.

Quick links

Overview

In this work, we present a simple approach for entity and relation extraction. Our approach contains three conponents:

  1. The entity model takes a piece of text as input and predicts all the entities at once.
  2. The relation model considers every pair of entities independently by inserting typed entity markers, and predicts the relation type for each pair.
  3. The approximation relation model supports batch computations, which enables efficient inference for the relation model.

Please find more details of this work in our paper.

Setup

Install dependencies

Please install all the dependency packages using the following command:

pip install -r requirements.txt

Download and preprocess the datasets

Our experiments are based on three datasets: ACE04, ACE05, and SciERC. Please find the links and pre-processing below:

  • ACE04/ACE05: We use the preprocessing code from DyGIE repo. Please follow the instructions to preprocess the ACE05 and ACE04 datasets.
  • SciERC: The preprocessed SciERC dataset can be downloaded in their project website.

Quick Start

The following commands can be used to download the preprocessed SciERC dataset and run our pre-trained models on SciERC.

# Download the SciERC dataset
wget http://nlp.cs.washington.edu/sciIE/data/sciERC_processed.tar.gz
mkdir scierc_data; tar -xf sciERC_processed.tar.gz -C scierc_data; rm -f sciERC_processed.tar.gz
scierc_dataset=scierc_data/processed_data/json/

# Download the pre-trained models (single-sentence)
mkdir scierc_models; cd scierc_models

# Download the pre-trained entity model
wget https://nlp.cs.princeton.edu/projects/pure/scierc_models/ent-scib-ctx0.zip
unzip ent-scib-ctx0.zip; rm -f ent-scib-ctx0.zip
scierc_ent_model=scierc_models/ent-scib-ctx0/

# Download the pre-trained full relation model
wget https://nlp.cs.princeton.edu/projects/pure/scierc_models/rel-scib-ctx0.zip
unzip rel-scib-ctx0.zip; rm -f rel-scib-ctx0.zip
scierc_rel_model=scierc_models/rel-scib-ctx0/

# Download the pre-trained approximation relation model
wget https://nlp.cs.princeton.edu/projects/pure/scierc_models/rel_approx-scib-ctx0.zip
unzip rel_approx-scib-ctx0.zip; rm -f rel_approx-scib-ctx0.zip
scierc_rel_model_approx=scierc_models/rel_approx-scib-ctx0/

cd ..

# Run the pre-trained entity model, the result will be stored in ${scierc_ent_model}/ent_pred_test.json
python run_entity.py \
    --do_eval --eval_test \
    --context_window 0 \
    --task scierc \
    --data_dir ${scierc_dataset} \
    --model allenai/scibert_scivocab_uncased \
    --output_dir ${scierc_ent_model}

# Run the pre-trained full relation model
python run_relation.py \
  --task scierc \
  --do_eval --eval_test \
  --model allenai/scibert_scivocab_uncased \
  --do_lower_case \
  --context_window 0\
  --max_seq_length 128 \
  --entity_output_dir ${scierc_ent_model} \
  --output_dir ${scierc_rel_model}
  
# Output end-to-end evaluation results
python run_eval.py --prediction_file ${scierc_rel_model}/predictions.json

# Run the pre-trained approximation relation model (with batch computation)
python run_relation_approx.py \
  --task scierc \
  --do_eval --eval_test \
  --model allenai/scibert_scivocab_uncased \
  --do_lower_case \
  --context_window 0\
  --max_seq_length 250 \
  --entity_output_dir ${scierc_ent_model} \
  --output_dir ${scierc_rel_model_approx} \
  --batch_computation

# Output end-to-end evaluation results
python run_eval.py --prediction_file ${scierc_rel_model_approx}/predictions.json

Entity Model

Input data format for the entity model

The input data format of the entity model is JSONL. Each line of the input file contains one document in the following format.

{
  # document ID (please make sure doc_key can be used to identify a certain document)
  "doc_key": "CNN_ENG_20030306_083604.6",

  # sentences in the document, each sentence is a list of tokens
  "sentences": [
    [...],
    [...],
    ["tens", "of", "thousands", "of", "college", ...],
    ...
  ],

  # entities (boundaries and entity type) in each sentence
  "ner": [
    [...],
    [...],
    [[26, 26, "LOC"], [14, 14, "PER"], ...], #the boundary positions are indexed in the document level
    ...,
  ],

  # relations (two spans and relation type) in each sentence
  "relations": [
    [...],
    [...],
    [[14, 14, 10, 10, "ORG-AFF"], [14, 14, 12, 13, "ORG-AFF"], ...],
    ...
  ]
}

Train/evaluate the entity model

You can use run_entity.py with --do_train to train an entity model and with --do_eval to evaluate an entity model. A trianing command template is as follow:

python run_entity.py \
    --do_train --do_eval [--eval_test] \
    --learning_rate=1e-5 --task_learning_rate=5e-4 \
    --train_batch_size=16 \
    --context_window {0 | 100 | 300} \
    --task {ace05 | ace04 | scierc} \
    --data_dir {directory of preprocessed dataset} \
    --model {bert-base-uncased | albert-xxlarge-v1 | allenai/scibert_scivocab_uncased} \
    --output_dir {directory of output files}

Arguments:

  • --learning_rate: the learning rate for BERT encoder parameters.
  • --task_learning_rate: the learning rate for task-specific parameters, i.e., the classifier head after the encoder.
  • --context_window: the context window size used in the model. 0 means using no contexts. In our cross-sentence entity experiments, we use --context_window 300 for BERT models and SciBERT models and use --context_window 100 for ALBERT models.
  • --model: the base transformer model. We use bert-base-uncased and albert-xxlarge-v1 for ACE04/ACE05 and use allenai/scibert_scivocab_uncased for SciERC.
  • --eval_test: whether evaluate on the test set or not.

The predictions of the entity model will be saved as a file (ent_pred_dev.json) in the output_dir directory. If you set --eval_test, the predictions (ent_pred_test.json) are on the test set. The prediction file of the entity model will be the input file of the relation model.

Relation Model

Input data format for the relation model

The input data format of the relation model is almost the same as that of the entity model, except that there is one more filed ."predicted_ner" to store the predictions of the entity model.

{
  "doc_key": "CNN_ENG_20030306_083604.6",
  "sentences": [...],
  "ner": [...],
  "relations": [...],
  "predicted_ner": [
    [...],
    [...],
    [[26, 26, "LOC"], [14, 15, "PER"], ...],
    ...
  ]
}

Train/evaluate the relation model:

You can use run_relation.py with --do_train to train a relation model and with --do_eval to evaluate a relation model. A trianing command template is as follow:

python run_relation.py \
  --task {ace05 | ace04 | scierc} \
  --do_train --train_file {path to the training json file of the dataset} \
  --do_eval [--eval_test] [--eval_with_gold] \
  --model {bert-base-uncased | albert-xxlarge-v1 | allenai/scibert_scivocab_uncased} \
  --do_lower_case \
  --train_batch_size 32 \
  --eval_batch_size 32 \
  --learning_rate 2e-5 \
  --num_train_epochs 10 \
  --context_window {0 | 100} \
  --max_seq_length {128 | 228} \
  --entity_output_dir {path to output files of the entity model} \
  --output_dir {directory of output files}

Aruguments:

  • --eval_with_gold: whether evaluate the model with the gold entities provided.
  • --entity_output_dir: the output directory of the entity model. The prediction files (ent_pred_dev.json or ent_pred_test.json) of the entity model should be in this directory.

The prediction results will be stored in the file predictions.json in the folder output_dir, and the format will be almost the same with the output file from the entity model, except that there is one more field "predicted_relations" for each document.

You can run the evaluation script to output the end-to-end performance (Ent, Rel, and Rel+) of the predictions.

python run_eval.py --prediction_file {path to output_dir}/predictions.json

Approximation relation model

You can use the following command to train an approximation model.

python run_relation_approx.py \
 --task {ace05 | ace04 | scierc} \
 --do_train --train_file {path to the training json file of the dataset} \
 --do_eval [--eval_with_gold] \
 --model {bert-base-uncased | allenai/scibert_scivocab_uncased} \
 --do_lower_case \
 --train_batch_size 32 \
 --eval_batch_size 32 \
 --learning_rate 2e-5 \
 --num_train_epochs 10 \
 --context_window {0 | 100} \
 --max_seq_length {128 | 228} \
 --entity_output_dir {path to output files of the entity model} \
 --output_dir {directory of output files}

Once you have a trained approximation model, you can enable efficient batch computation during inference with --batch_computation:

python run_relation_approx.py \
 --task {ace05 | ace04 | scierc} \
 --do_eval [--eval_test] [--eval_with_gold] \
 --model {bert-base-uncased | allenai/scibert_scivocab_uncased} \
 --do_lower_case \
 --eval_batch_size 32 \
 --context_window {0 | 100} \
 --max_seq_length 250 \
 --entity_output_dir {path to output files of the entity model} \
 --output_dir {directory of output files} \
 --batch_computation

Note: the current code does not support approximation models based on ALBERT.

Pre-trained Models

We release our pre-trained entity models and relation models for ACE05 and SciERC datasets.

Note: the performance of the pre-trained models might be slightly different from the reported numbers in the paper, since we reported the average numbers based on multiple runs.

Pre-trained models for ACE05

Entity models:

Relation models:

Performance of pretrained models on ACE05 test set:

  • BERT (single)
NER - P: 0.890260, R: 0.882944, F1: 0.886587
REL - P: 0.689624, R: 0.652476, F1: 0.670536
REL (strict) - P: 0.664830, R: 0.629018, F1: 0.646429
  • BERT-approx (single)
NER - P: 0.890260, R: 0.882944, F1: 0.886587
REL - P: 0.678899, R: 0.642919, F1: 0.660419
REL (strict) - P: 0.651376, R: 0.616855, F1: 0.633646
  • ALBERT (single)
NER - P: 0.900237, R: 0.901388, F1: 0.900812
REL - P: 0.739901, R: 0.652476, F1: 0.693444
REL (strict) - P: 0.698522, R: 0.615986, F1: 0.654663
  • BERT (cross)
NER - P: 0.902111, R: 0.905405, F1: 0.903755
REL - P: 0.701950, R: 0.656820, F1: 0.678636
REL (strict) - P: 0.668524, R: 0.625543, F1: 0.646320
  • BERT-approx (cross)
NER - P: 0.902111, R: 0.905405, F1: 0.903755
REL - P: 0.684448, R: 0.657689, F1: 0.670802
REL (strict) - P: 0.659132, R: 0.633362, F1: 0.645990
  • ALBERT (cross)
NER - P: 0.911111, R: 0.905953, F1: 0.908525
REL - P: 0.748521, R: 0.659427, F1: 0.701155
REL (strict) - P: 0.723866, R: 0.637706, F1: 0.678060

Pre-trained models for SciERC

Entity models:

Relation models:

Performance of pretrained models on SciERC test set:

  • SciBERT (single)
NER - P: 0.667857, R: 0.665875, F1: 0.666865
REL - P: 0.491614, R: 0.481520, F1: 0.486515
REL (strict) - P: 0.360587, R: 0.353183, F1: 0.356846
  • SciBERT-approx (single)
NER - P: 0.667857, R: 0.665875, F1: 0.666865
REL - P: 0.500000, R: 0.453799, F1: 0.475780
REL (strict) - P: 0.376697, R: 0.341889, F1: 0.358450
  • SciBERT (cross)
NER - P: 0.676223, R: 0.713947, F1: 0.694573
REL - P: 0.494797, R: 0.536961, F1: 0.515017
REL (strict) - P: 0.362346, R: 0.393224, F1: 0.377154
  • SciBERT-approx (cross)
NER - P: 0.676223, R: 0.713947, F1: 0.694573
REL - P: 0.483366, R: 0.507187, F1: 0.494990
REL (strict) - P: 0.356164, R: 0.373717, F1: 0.364729

Bugs or Questions?

If you have any questions related to the code or the paper, feel free to email Zexuan Zhong ([email protected]). If you encounter any problems when using the code, or want to report a bug, you can open an issue. Please try to specify the problem with details so we can help you better and quicker!

Citation

If you use our code in your research, please cite our work:

@inproceedings{zhong2021frustratingly,
   title={A Frustratingly Easy Approach for Entity and Relation Extraction},
   author={Zhong, Zexuan and Chen, Danqi},
   booktitle={North American Association for Computational Linguistics (NAACL)},
   year={2021}
}
Owner
Princeton Natural Language Processing
Princeton Natural Language Processing
A pytorch-based deep learning framework for multi-modal 2D/3D medical image segmentation

A 3D multi-modal medical image segmentation library in PyTorch We strongly believe in open and reproducible deep learning research. Our goal is to imp

Adaloglou Nikolas 1.2k Dec 27, 2022
Adaptive Attention Span for Reinforcement Learning

Adaptive Transformers in RL Official implementation of Adaptive Transformers in RL In this work we replicate several results from Stabilizing Transfor

100 Nov 15, 2022
Official implementation of "StyleCariGAN: Caricature Generation via StyleGAN Feature Map Modulation" (SIGGRAPH 2021)

StyleCariGAN in PyTorch Official implementation of StyleCariGAN:Caricature Generation via StyleGAN Feature Map Modulation in PyTorch Requirements PyTo

PeterZhouSZ 49 Oct 31, 2022
Location-Sensitive Visual Recognition with Cross-IOU Loss

The trained models are temporarily unavailable, but you can train the code using reasonable computational resource. Location-Sensitive Visual Recognit

Kaiwen Duan 146 Dec 25, 2022
A fast implementation of bss_eval metrics for blind source separation

fast_bss_eval Do you have a zillion BSS audio files to process and it is taking days ? Is your simulation never ending ? Fear no more! fast_bss_eval i

Robin Scheibler 99 Dec 13, 2022
pytorch implementation of trDesign

trdesign-pytorch This repository is a PyTorch implementation of the trDesign paper based on the official TensorFlow implementation. The initial port o

Learn Ventures Inc. 41 Dec 29, 2022
TransferNet: Learning Transferrable Knowledge for Semantic Segmentation with Deep Convolutional Neural Network

TransferNet: Learning Transferrable Knowledge for Semantic Segmentation with Deep Convolutional Neural Network Created by Seunghoon Hong, Junhyuk Oh,

42 Jun 29, 2022
Hierarchical Memory Matching Network for Video Object Segmentation (ICCV 2021)

Hierarchical Memory Matching Network for Video Object Segmentation Hongje Seong, Seoung Wug Oh, Joon-Young Lee, Seongwon Lee, Suhyeon Lee, Euntai Kim

Hongje Seong 72 Dec 14, 2022
A simple implementation of Kalman filter in single object tracking

kalman-filter-in-single-object-tracking A simple implementation of Kalman filter in single object tracking https://www.bilibili.com/video/BV1Qf4y1J7D4

130 Dec 26, 2022
Python implementation of "Multi-Instance Pose Networks: Rethinking Top-Down Pose Estimation"

MIPNet: Multi-Instance Pose Networks This repository is the official pytorch python implementation of "Multi-Instance Pose Networks: Rethinking Top-Do

Rawal Khirodkar 57 Dec 12, 2022
Self-Supervised Speech Pre-training and Representation Learning Toolkit.

What's New Sep 2021: We host a challenge in AAAI workshop: The 2nd Self-supervised Learning for Audio and Speech Processing! See SUPERB official site

s3prl 1.6k Jan 08, 2023
Enabling dynamic analysis of Legacy Embedded Systems in full emulated environment

PENecro This project is based on "Enabling dynamic analysis of Legacy Embedded Systems in full emulated environment", published on hardwear.io USA 202

Ta-Lun Yen 10 May 17, 2022
Official code for "Stereo Waterdrop Removal with Row-wise Dilated Attention (IROS2021)"

Stereo-Waterdrop-Removal-with-Row-wise-Dilated-Attention This repository includes official codes for "Stereo Waterdrop Removal with Row-wise Dilated A

29 Oct 01, 2022
Generative Query Network (GQN) in PyTorch as described in "Neural Scene Representation and Rendering"

Update 2019/06/24: A model trained on 10% of the Shepard-Metzler dataset has been added, the following notebook explains the main features of this mod

Jesper Wohlert 313 Dec 27, 2022
A repository for the updated version of CoinRun used to collect MUGEN, a multimodal video-audio-text dataset.

A repository for the updated version of CoinRun used to collect MUGEN, a multimodal video-audio-text dataset. This repo contains scripts to train RL agents to navigate the closed world and collect vi

MUGEN 11 Oct 22, 2022
Official PyTorch code for Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution (MANet, ICCV2021)

Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution (MANet, ICCV2021) This repository is the official PyTorc

Jingyun Liang 139 Dec 29, 2022
Old Photo Restoration (Official PyTorch Implementation)

Bringing Old Photo Back to Life (CVPR 2020 oral)

Microsoft 11.3k Dec 30, 2022
Official implementation of the paper DeFlow: Learning Complex Image Degradations from Unpaired Data with Conditional Flows

DeFlow: Learning Complex Image Degradations from Unpaired Data with Conditional Flows Official implementation of the paper DeFlow: Learning Complex Im

Valentin Wolf 86 Nov 16, 2022
I tried to apply the CAM algorithm to YOLOv4 and it worked.

YOLOV4:You Only Look Once目标检测模型在pytorch当中的实现 2021年2月7日更新: 加入letterbox_image的选项,关闭letterbox_image后网络的map得到大幅度提升。 目录 性能情况 Performance 实现的内容 Achievement

55 Dec 05, 2022
A keras-based real-time model for medical image segmentation (CFPNet-M)

CFPNet-M: A Light-Weight Encoder-Decoder Based Network for Multimodal Biomedical Image Real-Time Segmentation This repository contains the implementat

268 Nov 27, 2022