Closed kanghj closed 2 years ago
Hi Kang Hong, thanks for your interest in the paper. The visualization was made by manually organizing the data in powerpoint slides. Although, you may find the get_text
utility in data_process.ipynb notebook helpful to visualize top features of a row in a sparse csr matrix.
Thanks for the outstanding work on zero-shot XML!
The paper/supplementary material (https://nilesh2797.github.io/pubs/gupta21_supp.pdf) shows a nice UI that visualizes the top active instance label-feature pairs. It looks amazing and helpful to understand the predictions!
I would like to obtain the same analysis for my own work. Is the code to generate these visualizations (or to obtain the top voters) available?