This paper describes and evaluates a simple modification to the Brill Part-of-Speech Tagger. In its standard distribution the Brill Tagger allows manual assignment of a part-of-speech tag to a word prior to tagging. However, it may change it to another tag during processing. We suggest a change that guarantees that the pre-tag remains unchanged and ensures that it is used throughout the tagging process. Our method of guaranteed pre-tagging is appropriate when the tag of a word is known for certain, and is intended to help improve the accuracy of tagging by providing a reliable anchor or seed around which to tag.
|Original language||English (US)|
|Number of pages||10|
|Journal||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|State||Published - Dec 1 2003|