search

boudinfl / pke

Python Keyphrase Extraction module
GNU General Public License v3.0
1.57k stars 291 forks source link

speed up with new graph library #168

Closed shyambhu-mukherjee closed 2 years ago

shyambhu-mukherjee commented 3 years ago

I see that we are using the networkx project for building the graph. But several graphs have much less memory prints and higher speed such as igraph. would love to know what the authors of this lib think about this.

ygorg commented 3 years ago

Hi, thanks for this suggestion. If you are interested in this, you could set up an experiment to evaluate the speed and memory footprint of graph creation. I think the test should include at least TextRank and MultipartiteRank, and be tested on ACM (long documents), WWW and KP20k (if not too long) because they have a lot of document (see ake-datasets). The time to load the 2 libraries is also interesting to report.