Bug in DPSEG?

[connormayer]

Hello,

I'm working on a paper that compares different word segmentation algorithms, and wordseg has been extremely helpful. The wordseg documentation says that the DPSEG algorithm has a bug in it and "is not fully functional at present". I can't find any information about what the bug is/was and whether it has been fixed. I noticed in the source code that there's a function called _dpseg_bugfix to correct an issue with certain types of input. Is this the bug? I'm just hoping to confirm that DPSEG works properly before we report any results.

Thanks!

[mmmaat]

Hello,

I worked on that about 8 years ago and now I'm doing completely different things... I remember the bug was related to some numerical issues with the Monte Carlo process, C++ side, not Python side (which should just be a wrapper over the C++ code). The _dpseg_bugfix you mention is not related to that bug but is a fix to another (solved) bug.

[alecristia]

Hi Connor,
Apologies! To my knowledge, that bug still has not been fixed. Like Mathieu, I've also moved on to other topics... I wonder if the people who created dpseg may be more helpful?

[connormayer]

Hi Mathieu and Alex,

Thanks for the replies! I actually work with one of the authors of the dpseg paper, so I'll reach out to her.