How to train own model for NER?

Question

How to train own model for NER?

UNICASTERSOL opened this issue 7 years ago · comments

UNiCaster Solutions commented 7 years ago

Anyone knows how to train own NER model?

Sorry, the question should be:-
Anyone knows how to train own NER model? I have tried to train my own but I have reached to the stage whereby it run out memory. Anyone know what should I configure for optimum for CUTOFF and Iteration please?

Lucas Britz · Answer 1 · Tue Mar 10 2020 02:40:56 GMT+0800 (China Standard Time)

Did you got any solution @UNICASTERSOL ? I'm trying to get this working as well

MystaraCorvus · Answer 2 · Fri Oct 01 2021 03:27:37 GMT+0800 (China Standard Time)

I would love to know more. Did you receive any information?

Alex · Answer 3 · Mon Oct 11 2021 04:02:10 GMT+0800 (China Standard Time)

I haven't trained any NER model so far.
A good start would be to have a look at the original java project documentation.
From what I see in the code, you have to use the MaximumEntropyNameFinder class, and more specifically the TrainModel method. The input is a training file which should have a specific tagging with examples of the entities to detect (see NameFinderEventReader class to see how this training file is used).
Let me know if you manage to train your own model.