A Structured Self-attentive Sentence Embedding

Kaushal Shetty

Last update: Nov 28, 2022

Related tags

Text Data & NLP visualization deep-learning python3 pytorch classification attention attention-mechanism attention-model sentence-embeddings self-attentive-rnn self-attention attention-weights

Overview

Structured Self-attentive sentence embeddings

Implementation for the paper A Structured Self-Attentive Sentence Embedding, which was published in ICLR 2017: https://arxiv.org/abs/1703.03130 .

USAGE:

For binary sentiment classification on imdb dataset run : python classification.py "binary"

For multiclass classification on reuters dataset run : python classification.py "multiclass"

You can change the model parameters in the model_params.json file Other tranining parameters like number of attention hops etc can be configured in the config.json file.

If you want to use pretrained glove embeddings , set the use_embeddings parameter to "True" ,default is set to False. Do not forget to download the glove.6B.50d.txt and place it in the glove folder.

Implemented:

Classification using self attention
Regularization using Frobenius norm
Gradient clipping
Visualizing the attention weights

Instead of pruning ,used averaging over the sentence embeddings.

Visualization:

After training, the model is tested on 100 test points. Attention weights for the 100 test data are retrieved and used to visualize over the text using heatmaps. A file visualization.html gets saved in the visualization/ folder after successful training. The visualization code was provided by Zhouhan Lin (@hantek). Many thanks.

Below is a shot of the visualization on few datapoints.

Training accuracy 93.4% Tested on 1000 points with 90.2% accuracy

You might also like...

Sequence Modeling with Structured State Spaces

Structured State Spaces for Sequence Modeling This repository provides implementations and experiments for the following papers. S4 Efficiently Modeli

902 Jan 6, 2023

ANTLR (ANother Tool for Language Recognition) is a powerful parser generator for reading, processing, executing, or translating structured text or binary files.

13.6k Jan 5, 2023

Comments

Penalization term

In the original paper, the penalization term is equal to the Frobenius norm squared. In your implementation you did not consider the power 2. Did you try comparing both cases?

opened by moussaKam 6
Experiments

Hello dear @kaushalshetty ,

I'm trying to run your experiments, but I can get the same results. I made the same changes than webdizz, but the accuracy are always 50%. This default configuration should generate the 93.4% accuracy result?

Thanks

opened by heukirne 2
about classfication.py

when i run this sentence : classification_type = sys.argv[1] IndexError: list index out of range I don't know how to slove, can you tell me the solution?thanks

opened by Catherine-HFUT 1
How

Hello，thanks for your codes！I am trying to learn Self-Attention. But cant find the input files, Can you please give the sample input and other required files to run this example ?

opened by TingEn-Li 1