FISHQA ( Financial-Sentiment-Analysis-with-Hierarchical-Query-driven-Attention )

This is a Tensorflow implementation of Beyond Polarity: Interpretable Financial Sentiment Analysis with Hierarchical Query-driven Attention

Requirements

• python 3.6.1
• Tensorflow 1.11.0
• jieba 0.39

Code Introduction

Step 1: Preprocess data

python preprocess_train.py 
python preprocess_test.py

Preprocess training dataset/test dataset. Remember to modify the dictionary, fiterwords based on your own datasets.

Step 2: Training model

python train.py

cd FISHQA/code

Set params based on your own datasets and train you own model

Step 3: Test model

python test.py 

Step 4: Simple attention visualization

python view.py 

Data Introduction

• Modify your own queries(FISHQA/Query) based on your own datasets and prior knowledge. Each query can be manually decided.
• Notice that under folder temp/ is a subset of our preprocessed data.
• As the our dataset is private, we cannot release it. We put two raw samples in folder train_data and test_data individually.
• Under folder dictionary/, there are some extra dictionaries summarized by professional for Chinese financial news.