Thanks for the'hard method', when will you relase the soft version ? about ner_incomplete_annotation

[allanj]

Hi @hvingelby , I'm still in the middle of the integration. The early ACL deadline this year prevents me from doing so in a short time. But I will definitely make it useable right after the ACL deadline (on 9th Dec. 2019).

In the meanwhile, I would suggest you use the Dynet version. Do let me know if you run into any issues. I would love to help.

[allanj]

I'm pushing the soft verison to the master branch. Feel free to check and ask any questions

[allanj]

I'm on holiday at the moment and will finish it probably by 24 Sept. In the meanwhile, if you need to run the implementation immediately, you can check out the dynet version at previous commit https://github.com/allanj/ner_incomplete_annotation/tree/aa20c015b3f373ac4a1893e629ac8f2dd137faab or I can help run experiments with the dynet version.

[hvingelby]

Hi @allanj! What is the status of integrating the soft approach into the pytorch version?
I would really like to try out your approach on a dataset I have, but if the pytorch version is to far in the future I might try out the dynet approach instead 🙂

[yangshao]

@allanj Do you have any plan when to release the soft version? thanks

[allanj]

Guys, sorry for the pretty delayed late response. The soft version will be released by the end of this week. Thanks 🙂

[allanj]

Though I haven't released yet, I want to let you guys know that I have implemented the soft variant in the adding-soft-model branch for which you might want to check it out. I'm running the large-scale experiments right now to test the performance and will push to master branch if there is no problem.