morenolaquatra / auto-scientific-annotation

This repository contains the annotated citations for the ScisummNet dataset (https://cs.stanford.edu/~myasu/projects/scisumm_net/) using the automatic procedure proposed in the paper https://doi.org/10.1007/s11192-020-03532-3

Hi, I'm Moreno La Quatra

I'm a fellow researcher at STMLab.

• I’m interested in 📝 NLP, 🔉 Audio processing and 🖼️ Multimodal Learning.
• If you are a master student 👨‍🎓 👩‍🎓 interested in some of the fields above, you may want to take a look to this repo.
• I'm looking to collaborate on the analysis of spoken content 🗣️, get in touch if you are interested!
• 📫 Reach me by mail or on Twitter.

auto-scientific-annotation's Issues

Question about original paper

Hi, I read your paper《Exploiting pivot words to classify and summarize discourse facets of scientifc papers》and got many insights from this outstanding work. But I also got some questions about definition of the new Task "discourse facet summarization" and presented methods.

1. The goal is to extract a fixed-length, facet-specific summary of a reference paper. But what does the fixed-length means? the number of tokens or the number of sentences? And Is it an adjustable or predefined parameter?

2. The summary consists of a subset of rp's text spans pertaining to facet Fk. I didn't find specific definition of the "text spans"? Does the "text spans" means any sentence in full text of reference paper or cited text span found by task1A? I think it's an important question because we need to compute features for each text span in the regression method but the definition of "text span" is not clear for me.

3. We need supervised data in regression method. So I want to confirm whether the expected summary is the cited text spans in rp pertinent to a facet (You use the item "community summary" in the paper Experimental design). And the regression method is to predict y' by features x and make y' as close as possible to y (Rouge-L precision between text span and expected summary). Finally, we can use the regression model to predict overlap scores and choose text spans in proposed summary through scores.

4. I find that you try many regression models and show their performs in Results section. But I didn't find the detailed values for model parameters. Could these parameters be public?

If the related code can be open, I believe it will be of great help to the follow-up work!
Thanks for your wonderful work again and Hope you have a nice day!