taxogen icon indicating copy to clipboard operation
taxogen copied to clipboard

Cluster center or result of ranking function as new child node

Open jagol opened this issue 6 years ago • 0 comments

Reading the TaxoGen-paper, I thought new child nodes in the taxonomy were created by selecting the most representitive term for a topic (using the ranking algorithm which considers popularity and concentration). But reading through the code in main.py and cluster.py it seems to me that child nodes are created by selecting the center index for a cluster. Specifically the center index is found here: https://github.com/franticnerd/taxogen/blob/master/code/cluster.py#L34 the names of the center indices are returned here https://github.com/franticnerd/taxogen/blob/master/code/cluster.py#L70 and here they are used as new children nodes: https://github.com/franticnerd/taxogen/blob/master/code/main.py#L63

Did I maybe misunderstand the paper or the code? Or am I looking at the wrong version of the code? I am glad for any clarification.

jagol avatar Oct 18 '18 13:10 jagol