How to use with GigaWord Pretrained Text Summarization

[opennmt]

How to use with GigaWord Pretrained Text Summarization

[loretoparisi]

@guillaumekln hey I have to really thank you very much!

echo "indian prime minister p.v. narasimha rao 's promise of more autonomy for troubled kashmir and his plea for early state elections has sparked a violent reaction from provincial moslem and opposition parties ." | ./cli/translate --model /Users/loretoparisi/webservice/textsum_epoch7_14.69_release.t7 --beam_size 5
rao 's promise of more autonomy for kashmir

This is the first time I'm able to make a text summarization model (pre-trained) working. All the others (tensorflow, etc.), despite of all my (and not only) efforts, are too complicated/involes too much data not available/etc. to make it working as a simple testbed!

That's simply great.

[guillaumekln]

http://forum.opennmt.net/t/text-summarization-on-gigaword-and-rouge-scoring/85

[guillaumekln]

Looks like there is everything you need in this thread. What did you try so far?

[guillaumekln]

Depends on the input I guess. It works fine when using the Giga corpus as input.

[loretoparisi]

@guillaumekln Ok, so I'm going to download the training and evaluation data to test it, those available here: https://github.com/harvardnlp/sent-summary, right?

[guillaumekln]

Yes.

[loretoparisi]

And funny as well:

echo "The club isn't the best place to find a lover So the bar is where I go Me and my friends at the table doing shots Drinking fast" | ./cli/translate --model /Users/loretoparisi/webservice/textsum_epoch7_14.69_release.t7 --beam_size 10
go searching for love

👍

Be serious, what about the training part? The related paper A Neural Attention Model for Sentence Sum does not technically (of course, but sometimes they do) refers to a specific training etc. and I do not see here a specific tuturial for text sum.

Thank you.

[loretoparisi]

Thank you again, closing this.

[lchunleo]

may i know how to use the pretrained model for summarization? is there a tutorial on how to do so?