ASJP and DOLGO columns about autocogphylo HOT 5 CLOSED

You'll see, what's interesting: turchin is now better than edit distance!

from autocogphylo.

yes, will fix it asap, no deal, but need to recompute turchin distances, as they use the dolgo-information!

from autocogphylo.

Wow. . I pulled the data. Running OnlinePMI. I will update the results.

from autocogphylo.

You know that I use fixed thresholds now, based on the training in our PLOS paper. You should consider doing something similar, as I am convinced online-pmi will perform much better if you tweak the threshold according to some training data.

from autocogphylo.

In fact, we should not forget to look into the sound class issue, you know? Given how much of a difference it makes for the Turchin-method, I am quite sure that it will make similar differences for the other methods as well (even simply testing our three sound class models against a couple of datasets for different methods including one that uses all params, might be useful).

from autocogphylo.