System and Method for Korean Taging Using the Concatenation of Jamo and Sylable Embeding

摘要

The present invention relates to an apparatus and a method for analyzing a Korean morpheme by using the embedding-coupling of a letter and a syllable, which can determine an accurate word class even in a sentence with a frequent typing error by forming input by syllable and letter and using converted letter embedding. The apparatus includes: a letter unit embedding part performing initial, medial and final consonant embedding by letter; a syllable unit embedding part performing embedding by syllable; an input part coupling three initial/medial/final letter embedding, additionally coupling a syllable embedding to express one syllable as a vector, and providing the syllable as the input of Bi-LSTM-CRF; a learning part performing learning by using a backpropagation algorithm after conducting forward/backward steps of Bi-LSTM-CRF; and an output part using a Viterbi exploration algorithm to find an optimal tag string, and outputting a word class tag with symbols attached to indicate the start, middle and end of the word class.