https://github.com/celestialtaha/infersent

Implementation of Facebook's InferSent in Keras as well as a slightly improved version.
https://github.com/celestialtaha/infersent

facebook infersent keras nlp sentence-representations supervised-embeddings supervised-learning

Last synced: 19 days ago
JSON representation

Implementation of Facebook's InferSent in Keras as well as a slightly improved version.

• Host: GitHub
• URL: https://github.com/celestialtaha/infersent
• Owner: celestialtaha
• Created: 2021-04-09T05:12:40.000Z (about 5 years ago)
• Default Branch: main
• Last Pushed: 2021-04-09T05:26:08.000Z (about 5 years ago)
• Last Synced: 2025-07-21T19:27:01.383Z (11 months ago)
• Topics: facebook, infersent, keras, nlp, sentence-representations, supervised-embeddings, supervised-learning
• Language: Jupyter Notebook
• Homepage:
• Size: 2.34 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# InferSent

Facebook's InferSent Implemetation and Improvement using Keras framework

This is a Supervised algorithm of learning sentence representations, which has a hugely faster training time and comparable performance compared to unsupervised methods such as Skip-Thoughts and also capability of transferring to other non-sentiment based NLP tasks.

![image](https://github.com/tahasamavati/InferSent/blob/main/results.png)

Best Performing model is ESIM-FastText

There is no significant difference between using Glove or FastText Embeddings

The Proposed ESIM-FastText outperforms original model by 2.24%

---------------------------------------------------------------------------------

Taha Samavati - Analysis of final results