Code for First-K results

[loudly-soft]

Thanks for sharing the code and data!

Table 6 in the paper cites the Rouge results of First-3 (R1=39.41) being higher than LexRank (R1=38.27) and TextRank (R1=38.44). This seems a bit surprising. Would it be possible to share the code for First-K?

I've been trying to reproduce the First-3 results using the pre-processed test dataset but I'm not getting anywhere near R1=39.41 by following the method described in the paper:

For our dataset, First-k means the first k sentences from each source article will be concatenated as the summary.