NE type consistency in relation to nerd-kid

[tantikristanti]

The same mention can have different types.

One example can be found in the result of a text disambiguation service with the text PubMed_2. It shows that has the same mentions can have different of Type.

[kermitt2]

Hello this is done on purpose, as we don't know which type is the good one a priori, and the type of the base mention is not overwritten to the acronyms.

Basically I am waiting a mapping of the entities to the types (something like nerd-kid ;) ), and there is nothing particular to the fact it is acronym or not in this situation.

[kermitt2]

I suggest to either close it or rename/reopen it to broad NE type consistency in relation to nerd-kid.

[tantikristanti]

Responding to the Named Entity consistency issue #74, a new branch called "0.0.3-nerdKid" has been built for the purpose of developing a mapping of the entitites to the types.

Even though the works are still in progress, but to give some ideas of how it would look like, the results of a dedicated database containing the mapping of entities into the types (e.g. Person, Location, Organisation) have been "inserted" in the demo console of the branch as a field Predicted as.

The results of text disambiguation then can be seen as follows:

• some entities that are not recognized by entity-fishing became recognizable (the field Predicted as)
• some entities that are recognized but not consistent, like for example, "TAK1" and "CS" became consistent as "Other"

Further works are to ascertain whether the inclusion of the results of the entities <-> class types will give a better impact.

[tantikristanti]

For entity-fishing version 0.0.4, the NE type generated by nerdKid has been integrated in branch 0.0.4-withNerdKidTypeInDemo