SSJE

Code for A span-sharing joint extraction framework for harvesting aspect sentiment triplets.(Knowledge-Based Systems'2022)

Model Architecture

Requirments

Enviroment

python 3.6.1
numpy==1.17.4
tensorboardX==1.6
scikit-learn==0.25.0
torch==1.4.0
transformers
spacy==3.0.1
tpqm

use pip command

pip install -r requirements.txt

Notes

You can download the bert-base-cased from here

File Directory Tree

The directory tree of SSJE:

├─bert
│  └─base-uncased
├─data
│  ├─14lap
│  ├─14res
│  ├─15res
│  └─16res
├─layer
├─log
├─model
├─templates
├─trainer

Get Started

Datasets

download the ASTE-Dataset-V2 dataset from here or you can just use the data set that we've already processed.

Run

• Training and testing model effects on 2014 Restaurant

python train_triplet.py --dataset 14res --max_span_size 8