GithubHelp home page GithubHelp logo

Comments (2)

nicpopovic avatar nicpopovic commented on September 6, 2024 2

Hi David,

Thank you for your interest and your questions!

  1. Are the support documents for the cross-domain test set (comprising solely of SciERC samples) sampled from your Train+Dev set (62+16 relation types from DocRed)?

The support documents for the cross-domain test set are sampled from the cross-domain test set itself. Sampling support documents from the in-domain set would not be possible (at least for the task format we chose) because the support documents need to contain instances of the relations which are to be extracted from the query documents. Since there is no overlap1 between relation types in the in-domain sets and the cross-domain set, a task in which the support and query documents are sampled from separate sets is not possible for the two datasets.

  1. If the answer to question (1) is yes, is this not zero-shot with the setting that shot is defined as relation types trained/encountered?

The answer to (1) is no, but if it where yes, it would sound to me more like a zero-shot setting than a few-shot setting, yes. However, I am not sure the support documents would be useful/usable input for such a task, as they would contain information about neither the query domain nor the relation types which are to be extracted.

  1. Just to be certain, where are the support docs for the in-domain test set (16 RT from DocRED) sampled from? (Train+Dev set or in-domain test set)?

The support documents for the in-domain test set are sampled from the in-domain test set. In general, support and query documents will always be sampled from the same set. This is because the support and query documents need to be annotated with the same relation types. Note that during testing, the model does not perform any (persistent) learning on the test episodes.

I hope that I was able to answer your questions?

Nicholas

Footnotes

  1. As mentioned in the paper, there technically are 2 relation types which are contained in both DocRED and SciERC, but we remove these from the in-domain set.

from fredo.

davidleejy avatar davidleejy commented on September 6, 2024 2

These are very clear answers. Thank you for the elaboration, and congrats on the acceptance to NAACL.

from fredo.

Related Issues (4)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.