---
language:
- sci
- en
tags:
- classification
- similarity
license: mit
datasets:
- cord19
---

# Aspect-based Document Similarity for Research Papers

A `scibert-scivocab-uncased` model fine-tuned on the CORD-19 corpus as in [Aspect-based Document Similarity for Research Papers](https://arxiv.org/abs/2010.06395).

<img src="https://raw.githubusercontent.com/malteos/aspect-document-similarity/master/docrel.png">

See GitHub for more details: https://github.com/malteos/aspect-document-similarity

## Demo

<a href="https://colab.research.google.com/github/malteos/aspect-document-similarity/blob/master/demo.ipynb"><img src="https://camo.githubusercontent.com/52feade06f2fecbf006889a904d221e6a730c194/68747470733a2f2f636f6c61622e72657365617263682e676f6f676c652e636f6d2f6173736574732f636f6c61622d62616467652e737667" alt="Google Colab"></a>

You can try our trained models directly on Google Colab on all papers available on Semantic Scholar (via DOI, ArXiv ID, ACL ID, PubMed ID):

<a href="https://colab.research.google.com/github/malteos/aspect-document-similarity/blob/master/demo.ipynb"><img src="https://raw.githubusercontent.com/malteos/aspect-document-similarity/master/demo.gif" alt="Click here for demo"></a>