Self-Attentive-Tensorflow

Tensorflow implementation of A Structured Self-Attentive Sentence Embedding

You can read more about concept from this paper

Key Concept

Frobenius norm with attention

Usage

Download ag news dataset as below

$ tree ./data
./data
└── ag_news_csv
    ├── classes.txt
    ├── readme.txt
    ├── test.csv
    ├── train.csv
    └── train_mini.csv

and then

$ python train.py

Result

Accuracy 0.895

visualize without penalization

visualize with penalization

To-do list

• support multiple dataset

Notes

This implementation does not use pretrained GloVe or Word2vec.