The present invention relates to a method and system for constructing a named entity dictionary by unsupervised learning. The method includes: a categorized document data collection step of collecting document data on the internet and then extracting a category of the document data; a named entity usage pattern registration step of analyzing a sentence structure by a natural language processing procedure with respect to the document data, analyzing a usage pattern for each category for the named entity based on meaning of each category of the named entity registered in a named entity dictionary DB, and then registering an extracted official usage pattern for each category in the usage pattern DB for each category; and a named entity dictionary registration step of applying the official usage pattern, registered in the usage pattern DB for each category according to a category, to a usage pattern wherein an unregistered named entity is used in document data including the unregistered named entity among the document data, matching the unregistered named entity to the meaning, and then registering the unregistered named entity as a new named entity in the named entity dictionary DB. According to the present invention, meaning of a named entity for each category is figured out by comprehensively analyzing categorized document data. Therefore, reliability of recognizing a newly introduced named entity is increased, and, furthermore, change in meaning of the name entity can be effectively recognized.
展开▼