0

I am searching for an R package that implements MaxEnt for NLP tasks (i.e. for situations with large no. of parameters).

The requirement is to perform Named Entity Recognition on medical text. I got used to relying on CRF++ for sequence labelling jobs but for this particular entity recognition task at hand, I dont want the transition/label input.

I have checked Maxent Package - it creates a tf-idf representation which I dont want. I need to provide my features as combinations of previous and next words, pos tags, etc. Very similar to how one would use CRF++.

mejobhoot
  • 71
  • 8
  • This question is an extension of this discussion: https://stats.stackexchange.com/questions/17436/logistic-regression-with-lbfgs-solver that did not seem to have an answer for an R implementation. – mejobhoot May 04 '17 at 13:19
  • There is no decent R-native solution for named entity recognition. You can use `spacyr` which is an R wrapper for spaCy (in Python). It's not on cran but it works mostly fine on most platforms (so I believe as a developer of the package). https://github.com/kbenoit/spacyr – amatsuo_net May 04 '17 at 13:29
  • Thanks. I will try out spaCy and see how it goes. – mejobhoot May 07 '17 at 14:27

0 Answers0