Protein name tagging for biomedical annotation in text

We explore the use of morphological analysis as preprocessing for protein name tagging. Our method finds protein names by chunking based on a morpheme, the smallest unit determined by the morphological analysis. This helps to recognize the exact boundaries of protein names. Moreover, our morphological analyzer can deal with compounds. This offers a simple way to adapt name descriptions from biomedical resources for language processing. Using GENIA corpus 3.01, our method attains f-score of 70 points for protein molecule names, and 75 points for protein names including molecules, families and domains.