NATURAL LANGUAGE PROCESSING SYSTEM AND METHOD FOR WORD REPRESENTATIONS IN NATURAL LANGUAGE PROCESSING

摘要

The present invention relates to a natural language processing system and a word expression method in processing a natural language. A word expression method in a natural language process performed by a natural language processing system includes the steps of: a) providing a vocabulary dictionary dataset including a vocabulary including at least one word and pre-trained word embedding information for each word; b) when a vocabulary based on the vocabulary dictionary dataset is provided as input data, extracting subword information for words existing in the input data by using a word standard model, and calculating a word embedding vector based on the subword information; and c) replacing the pre-trained word embedding information with the calculated word embedding vector by matching the calculated word embedding vector with the pre-trained word embedding information of the corresponding word to learn a word expression for the corresponding word. The word standard model includes convolutional neural network-based convolution module that calculates lower word feature vectors by using the lower word information; and a highway network-based highway module that adaptively combines the lower word feature vectors calculated by the convolution module to calculate the word embedding vector of the corresponding word.