Large part-of-speech(pos) annotated corpus play an important role in many kinds of natural language processing. So, the annotated corpus requires very high accuracy and consistency. To build such accurate and consistent corpus, we often use manual tagging. But the manual tagging is very labor intensive and expensive. Furthermore, it is not easy to get consistent results from the human experts. The goal of this work is to develope an efficient tool for building accurate and a consistent pos annotated corpus with minimal human labor. The developed tool can help minimize the amount of the human labor and make the results consistent by using lexical rules. The lexical rules are acquired from human experts in the similar way of manual tagging and manual error correction. They are used to annotate the same word in the same context in the whole corpus.
展开▼