talha1503 / Get-To-The-Point-Summarization-with-Pointer-Generator-Networks

Pytorch implementation of Get To The Point: Summarization with Pointer-Generator Networks (2017) by Abigail See et al.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Get To The Point: Summarization with Pointer-Generator Networks

Pytorch implementation of Get To The Point: Summarization with Pointer-Generator Networks (2017) by Abigail See et al.

Model Description

  • LSTM based Sequence-to-Sequence model for Abstractive Summarization
  • Pointer mechanism for handling Out of Vocabulary (OOV) words See et al. (2017)
  • FastText used for creating embeddings over the dataset .

Model Architecture

Prerequisites

  • Pytorch
  • gensim
  • python 3

Data

Data set used : Kaggle:Amazon fine food reviews

Examples

Sentence: once more amazon was great the product is good for kids even though it has a little bit more sugar than needed
Predicted Summary: good as expected
Actual Summary: as expected

Sentence: this is an excellent tea for a breakfast tea or for the afternoon or evening it has a wonderful mellow flavor in the morning i like to brew it with earl grey to create a nice smooth blend it is a great way to start the day
Predicted Summary: great product tea
Actual Summary: wonderful anytime tea

Sentence: very tasty this mix makes a moist yummy cake we make our own cream cheese frosting with the chocolate cake and it is a winner
Predicted Summary: tasty
Actual Summary: versatile and yummy

Sentence: ca not complain taste good and had quick delivery it was my first time trying this tea out i usually drink the peppermint one but this gave me energy and sustained me throughout the day
Predicted Summary: very good good
Actual Summary: very good

Sentence: energy bites f ing rip your face off with molten lava energy br br infinity energy to the f ing max i ate a whole box once and threw a car at a baby br br f ing rave br br seriously these are great
Predicted Summary: love it
Actual Summary: maximum rave power

Sentence: looks good from the package but does not taste good at all hard crunchy freeze dried flavor br save your money guys
Predicted Summary: yuck
Actual Summary: yuck

Sentence: this is the best dog food their is because everything is very digestible and when the dog does digest it all digested because they use all the nutrition from it so it is healther and the best out their far as i am think
Predicted Summary: best dog food
Actual Summary: natures logic venisen

Sentence: great coffee br excellent service br best way to buy k cups br stock up before coffee prices go up again
Predicted Summary: great coffee
Actual Summary: great coffee excellent service

Sentence: wonderful flavor would purchase this blend of coffee again light flavor not bitter at all and price was great the best i found anywhere
Predicted Summary: great flavor
Actual Summary: wolfgang puck k cup breakfast in bed

Sentence: no wonder they were so cheap they do not work very well in my machine because of the biodegradable packaging i would not buy them again
Predicted Summary: disappointed
Actual Summary: ethical coffee nespresso capsules

Sentence: i bought these from a large chain pet store after reading the reviews i checked the bag made in china i threw the whole bag away i wish i would have read the reviews first
Predicted Summary: do not buy
Actual Summary: do not buy

Sentence: i love these gums they are not as cloyingly sweet as american gummies and have a lot more fruit flavor the only problem is that i ca not eat just one
Predicted Summary: a addictive
Actual Summary: danger highly addictive

Sentence: the pepper plant habanero extra hot california style hot pepper sauce 10 oz has great flavor as all the pepper plants do i just love it it is a bit pricey but worth it
Predicted Summary: great seasoning
Actual Summary: wonderful love it


Loss for 75k iterations



Contributors:


References

About

Pytorch implementation of Get To The Point: Summarization with Pointer-Generator Networks (2017) by Abigail See et al.


Languages

Language:Jupyter Notebook 100.0%