However this is a model developed in Java, with principles that are different from supervised learning. It seems it doesn't fit with the logic in place to upload models to Dynalab. To make interfacing with it easier, I have added two interfaces to access it:
Would it be possible to have it integrated in some way to DynaBench? I am fully available to develop this alternative way to submit models. Thank you! Clement
Hi! I have developed Umigon, a lexicon-based model for sentiment analysis on social media texts. It is open sourced there. It ranked best out of 24 solutions on 3-class sentiment categorization, a while ago. I have refined it over the years so I'd expect its accuracy to have improved. I would naturally love to have evaluated on Dynasent.
However this is a model developed in Java, with principles that are different from supervised learning. It seems it doesn't fit with the logic in place to upload models to Dynalab. To make interfacing with it easier, I have added two interfaces to access it:
Would it be possible to have it integrated in some way to DynaBench? I am fully available to develop this alternative way to submit models. Thank you! Clement