theodoric008/senet-for-weakly-supervised-relation-extraction

This is the implementation of my paper: SENet for Weakly-Supervised Relation Extraction (CSAI 2018)

Click here for pdf draft: paper_draft

Accepted link: // todo

How to train?

1. unzip zipfile in data/ (the dataset is too large that you'd better download from here: https://github.com/darrenyaoyao/ResCNN_RelationExtraction/tree/master/data )
2. in cmd:

python3 train.py

and test result will be saved to temp/ in format of pkl file

How to eval?

python3 eval.py

How to plot and compare with other models?

you need to fill in the pkl file path in plot script, and run

cd plot/
python3 plot_compare_with_other_model.py
python3 metric.py

Model structure

Best result(epoch ~= 170)

Prerequisits

1. Tensorflow-gpu==1.4.0
2. sklearn, tflearn, nltk, numpy
3. Python3

Other models for RE and some helpful repos

PCNN + ATT

ResCNN-9

Linguistic_adversity

About me

Master candidate from PRIS, BUPT.

Email: [email protected]

senet-for-weakly-supervised-relation-extraction's Issues

请问plot.py中的ResCNN绘图数据3920.pkl在哪里获取？

# ResCNN 9

# Fill in the path of the model that you wanna compare with

pkls = '3920.pkl'

publish

may I ask you where your paper publish ,thank you

TypeError: unhashable type: 'list'

gpuws@gpuws32g:~/ub16_prj/SENet-for-Weakly-Supervised-Relation-Extraction$ python3.5 train.py
2018-12-03 20:38:09.902655: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA
2018-12-03 20:38:10.010289: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:892] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-12-03 20:38:10.010727: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Found device 0 with properties:
name: GeForce GTX 1080 Ti major: 6 minor: 1 memoryClockRate(GHz): 1.6575
pciBusID: 0000:01:00.0
totalMemory: 10.91GiB freeMemory: 8.59GiB
2018-12-03 20:38:10.010769: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: GeForce GTX 1080 Ti, pci bus id: 0000:01:00.0, compute capability: 6.1)

Parameters:
ALLOW_SOFT_PLACEMENT=True
BATCH_SIZE=64
DROPOUT_KEEP_PROB=0.5
EMBEDDING_DIM=50
FILTER_SIZES=3
L2_REG_LAMBDA=0.0001
LOG_DEVICE_PLACEMENT=False
NUM_EPOCHS=300
NUM_FILTERS=128
SEQUENCE_LENGTH=100

WordTotal= 114043
Word dimension= 50
RelationTotal: 53
Start loading training data.

Start loading testing data.

train set and test set size are:
570088 96678
Finish randomize data
Start Training
2018-12-03 20:38:34.521647: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: GeForce GTX 1080 Ti, pci bus id: 0000:01:00.0, compute capability: 6.1)
Initialize variables.
Batch data
num_epoch 1 epoch_step 1 / 8908 loss 3.94999, acc 0.03125
num_epoch 1 epoch_step 2 / 8908 loss 2.20825, acc 0.734375
num_epoch 1 epoch_step 3 / 8908 loss 2.50987, acc 0.65625
num_epoch 1 epoch_step 4 / 8908 loss 2.26494, acc 0.625
num_epoch 1 epoch_step 5 / 8908 loss 1.84646, acc 0.578125
num_epoch 1 epoch_step 6 / 8908 loss 1.58758, acc 0.6875
num_epoch 1 epoch_step 7 / 8908 loss 1.43806, acc 0.71875
num_epoch 1 epoch_step 8 / 8908 loss 1.4291, acc 0.765625
num_epoch 1 epoch_step 9 / 8908 loss 1.9987, acc 0.65625
num_epoch 1 epoch_step 10 / 8908 loss 1.43472, acc 0.703125
num_epoch 1 epoch_step 11 / 8908 loss 1.66532, acc 0.6875
num_epoch 1 epoch_step 12 / 8908 loss 1.65045, acc 0.640625
num_epoch 1 epoch_step 13 / 8908 loss 1.86547, acc 0.59375
num_epoch 1 epoch_step 14 / 8908 loss 1.5699, acc 0.640625
num_epoch 1 epoch_step 15 / 8908 loss 1.97835, acc 0.625
num_epoch 1 epoch_step 16 / 8908 loss 1.57536, acc 0.71875
num_epoch 1 epoch_step 17 / 8908 loss 1.79578, acc 0.59375
num_epoch 1 epoch_step 18 / 8908 loss 1.44287, acc 0.734375
num_epoch 1 epoch_step 19 / 8908 loss 1.21793, acc 0.703125
num_epoch 1 epoch_step 20 / 8908 loss 1.39004, acc 0.765625
num_epoch 1 epoch_step 21 / 8908 loss 1.14212, acc 0.796875
num_epoch 1 epoch_step 22 / 8908 loss 0.91511, acc 0.828125
num_epoch 1 epoch_step 23 / 8908 loss 2.20525, acc 0.625
num_epoch 1 epoch_step 24 / 8908 loss 1.63561, acc 0.71875
num_epoch 1 epoch_step 25 / 8908 loss 1.36389, acc 0.734375
num_epoch 1 epoch_step 26 / 8908 loss 1.01296, acc 0.84375
num_epoch 1 epoch_step 27 / 8908 loss 1.24955, acc 0.78125
num_epoch 1 epoch_step 28 / 8908 loss 1.52108, acc 0.6875
num_epoch 1 epoch_step 29 / 8908 loss 1.21105, acc 0.6875
num_epoch 1 epoch_step 30 / 8908 loss 1.24232, acc 0.765625
num_epoch 1 epoch_step 31 / 8908 loss 1.69791, acc 0.65625
num_epoch 1 epoch_step 32 / 8908 loss 1.56406, acc 0.6875
num_epoch 1 epoch_step 33 / 8908 loss 1.4502, acc 0.6875
num_epoch 1 epoch_step 34 / 8908 loss 1.43388, acc 0.75
num_epoch 1 epoch_step 35 / 8908 loss 1.20509, acc 0.703125
Traceback (most recent call last):
File "train.py", line 160, in
batch = data_aug(batch)
File "train.py", line 91, in data_aug
data_item.words = aug(data_item)
File "train.py", line 78, in aug
pkl_dict, random_lin_adv_prob)
File "/home/gpuws/ub16_prj/SENet-for-Weakly-Supervised-Relation-Extraction/sentence_aug.py", line 36, in random_lin_adv_noise
aug_sentences = pkl_dict[entity_pair][sentence]
TypeError: unhashable type: 'list'
gpuws@gpuws32g:~/ub16_prj/SENet-for-Weakly-Supervised-Relation-Extraction$

Where is

missing data/re ?

ub16hp@UB16HP:~/ub16_prj/SENet-for-Weakly-Supervised-Relation-Extraction$ python3 train.py
2018-10-31 11:35:16.509910: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA
2018-10-31 11:35:16.577874: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:892] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-10-31 11:35:16.578199: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Found device 0 with properties:
name: GeForce GTX 950M major: 5 minor: 0 memoryClockRate(GHz): 1.124
pciBusID: 0000:01:00.0
totalMemory: 3.95GiB freeMemory: 3.56GiB
2018-10-31 11:35:16.578221: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: GeForce GTX 950M, pci bus id: 0000:01:00.0, compute capability: 5.0)

Parameters:
ALLOW_SOFT_PLACEMENT=True
BATCH_SIZE=64
DROPOUT_KEEP_PROB=0.5
EMBEDDING_DIM=50
FILTER_SIZES=3
L2_REG_LAMBDA=0.0001
LOG_DEVICE_PLACEMENT=False
NUM_EPOCHS=300
NUM_FILTERS=128
SEQUENCE_LENGTH=100

WordTotal= 114043
Word dimension= 50
Traceback (most recent call last):
File "train.py", line 50, in
datamanager = DataManager(FLAGS.sequence_length)
File "/home/ub16hp/ub16_prj/SENet-for-Weakly-Supervised-Relation-Extraction/util/DataManager.py", line 21, in init
self.load_relations()
File "/home/ub16hp/ub16_prj/SENet-for-Weakly-Supervised-Relation-Extraction/util/DataManager.py", line 47, in load_relations
relation_data = list(open("data/RE/relation2id.txt", encoding='utf-8').readlines())
FileNotFoundError: [Errno 2] No such file or directory: 'data/RE/relation2id.txt'
Exception ignored in: <bound method BaseSession.del of <tensorflow.python.client.session.Session object at 0x7f5e1875ca90>>
Traceback (most recent call last):
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 696, in del
TypeError: 'NoneType' object is not callable
ub16hp@UB16HP:~/ub16_prj/SENet-for-Weakly-Supervised-Relation-Extraction$