pedrodiamel / pytorch-paraphraser

pytorch paraphraser generate

Pytorch Paraphrase Generation

This project providers users the ability to do paraphrase generation for sentences through a clean and simple API.

TODO

• Training framework (train|visualization|test)
• Evaluate methods
• Module for test/api (UI Flask module)
• Manifold implementation
• Pip installable package
• Docker configurate packaging
• Readme and doc

Datasets

• para-nmt-5m
• Quora
• snli
• PPDB
• Semeval

Models

• Attention NMT
• Embedded Representation
• Manifold

Results

Server	Name	Dataset	Bleu	Description
@Moto	AttNMT	CMDS	0.7163	Train: DB(para-nmt-5m-short+cmd)
@Moto	AttNMT	CMDS+G	0.1351	Train: DB(para-nmt-5m-shoet+cmd)
@Learn	TRP+GRU	CMDS+G	0.0047	Train: DB(para-nmt-5m-shoet+cmd)

Reference

• https://arxiv.org/pdf/1706.01847.pdf
• http://www.cs.cmu.edu/~jwieting/wieting2017Millions.pdf
• https://arxiv.org/pdf/1711.05732.pdf
• https://www.aaai.org/ocs/index.php/AAAI/AAAI18/paper/view/16353/16062
• https://arxiv.org/pdf/1808.04364.pdf
• https://arxiv.org/pdf/1803.01465.pdf
• http://alborz-geramifard.com/workshops/nips18-Conversational-AI/Papers/18convai-Neural%20Machine%20Translation.pdf
• https://arxiv.org/pdf/1901.02998.pdf
• https://arxiv.org/pdf/1804.06059.pdf

Repos

• https://github.com/kefirski?tab=repositories
• https://github.com/jhclark/multeval
• https://github.com/vsuthichai/paraphraser
• https://github.com/jwieting/para-nmt-50m