search

JRC1995 / Abstractive-Summarization

Implementation of abstractive summarization using LSTM in the encoder-decoder architecture with local attention.
MIT License
167 stars 59 forks source link

where does the abstraction come from? #12

Closed bhomass closed 5 years ago

bhomass commented 5 years ago

Hi I looked at the amazon fine food reviews data. It doesn't seem the review articles are accompanied by summaries. Are they?

Is the abstractive summary part based on any paper, such as Abigail see's pointer-generator? I only see references to lstm papers.

There is no mention of how you dealt with OOV words. Can you explain how you handled that?

JRC1995 commented 5 years ago

It doesn't seem the review articles are accompanied by summaries. Are they?

They are. Check data column descriptions: https://www.kaggle.com/snap/amazon-fine-food-reviews#Reviews.cs

Summary Brief summary of the review Text Text of the review

Is the abstractive summary part based on any paper, such as Abigail see's pointer-generator? I only see references to lstm papers.

It's a basic LSTM based encoder-decoder architecturer with (local) attention. I haven't used any pointer or copy mechanism. It does also have residual recurrent attention (RRA) or something named like that (there should be references to the papers in the notebook) but I am suspicious of that mechanism now a days.

There is no mention of how you dealt with OOV words. Can you explain how you handled that?

They are just replaced by a special tag '' or something like that. Pointer mechanism is a better way to deal with OOV words that appears in source text, but I didn't know of Pointer mechanism when I initially created the repo.