Most Frequent Sense documentation

[jnehring]

FREME NER has the most frequent sense (MFS) disambiguation method. To my knowledge this requires certain values to be initialized in FREME NER that determine that one word sense is more frequent then another.

Does the documentation explain how to initialize this data? I dont think so.

If not, please add a section to the freme ner article how to initialise the MFS values.

What happens when we use MFS and the counts are not initialized when FREME NER has to choose between two senses with the same surface form? Does it choose the first in the list?

[m1ci]

This is valid for DBpedia, since such data is not available for other datasets.
@sandroacoelho can you please extend the FREME NER article and explain

1. what data is needed for most frequent sense
2. how it can be loaded in Solr

@jnehring

 What happens when we use MFS and the counts are not initialized when FREME NER has to choose between two senses with the same surface form?

Its by Solr. If the mention is same with two surface forms which are same but for different entities, then I would say that Solr ranks them randomly - without any specific preference for one over another.

Does it choose the first in the list?

yes.

[sandroacoelho]

Hi @m1ci : sure. Let's do that