This project will cover different experiments for relation extraction.
Experiment 1:
Simple CNN - 3 conv 3 pool layers fully connected network in the end
Glove vectors
NYT Corpus
Experiment 2:
Capsule Network Implementation [capsnet.py].
Experiment 3:
Better, faster and lighter Capsule Network [ capsNetv2.py]
Experiment 4:
Adding 1*1 convolutions in Experiment 1
Experiment 5:
Hierarchial networks with Attention for Relation Extraction with zero state
for each sentence
Exploring seq2seq models with attention
============================
Use language model
Experiment 6:
Increasing layers in Capsule network. Find new way to calc loss. Remove NAN and INF.
Experiment 7:
Memory networks
Experiment 8:
GAN for Relation Extraction
Experiment 9:
RL for Relation Extraction
Other Experiments:
Combining 2 graphs and keeping only 1 trainable