I am looking to apply an open domain ontology and a part of speech tagger to disambiguate my corpus of natural language utterances (trained data). I found that Language Model (LM) rendering is limited in matter of accuracy.