1- I have to build a separate case for each model, for example. for a non-person to build a body where only tags are present? IMO yes ... however, it is possible that the model contains several types of names. If you save them separately, you can more easily update and iteratively improve models for given names, especially if they are large models.
2- Is it possible to mark a person, location and organization in one building and use it to train one model that can extract all three types of entities? yes, itβs possible, but if you plan to build on each type of name and improve the models, keeping them separate was easier for me.
3 - is there a resource where I can find more about the algorithm used in the OpenNLP Name finder module? The best thing was to do this, pull out the source and execute the code with some real data ... it is based on maximum entropy.
markg source share