yixinL7
commented
3 years ago Hi Subin, thank you for your interest in our work!
Is there any reason that pre-trained Refactor is not suitable as base model in re-ranking?
For the re-ranking setting, we are using abstractive models to sample different candidate summaries. Refactor is not suitable as a base system in this setting since it is an extractive model.
Parameter sharing between base and meta models is not directly associated with performance improvement. Instead, it is a feature of our refactor model which allows it to work both as a base model and a meta model. In $5.6, refactor serves as both a strong base model and the meta model. Other summarization models with different architectures will not be able to do this. The advantage of this feature is that we simultaneously get a pre-trained meta model and a strong base model.
Please let me know if you have more questions!
aqaqsubin
Dear Yixin Liu,
I really enjoyed your paper on 2-stage learning framework utilizing the complementarity of different text summarization systems.
I have some questions about paper.
On your paper, you say as follows. Sharing a set of parameters between base and meta system can alleviate the "Base-Meta learning gap". And, pretrain-then-finetune paradiam can mitigate "Train-Test distribution gap".
But i think, there is no experiment indicating that sharing parameters between base-meta model can boost performance. In $5.5 and $5.8 experiment, pre-trained Refactor is only used as meta learner. (base model is BART, GSum and PEGASUS) In $5.6 experiment, pre-trained Refactor is also used as base model, but there is no parameter sharing between the other base model and meta model.
Is there any reason that pre-trained Refactor is not suitable as base model in re-ranking? And is there performance difference when no parameter sharing with meta model (i.e. Refactor only used as meta model in multi-system) and when sharing parameter between meta and only one base model (in $5.6)?
please correct me if I got it wrong.
Thanks in advance.