APPARATUS AND METHOD FOR EMBEDDING SENTENCE FEATURE VECTOR
An apparatus and a method for embedding a sentence feature vector are disclosed. A sentence feature vector embedding apparatus in accordance with an embodiment of the present invention includes: a sentence acquisition unit configured to acquire a first sentence and a second sentence including one or...
Saved in:
Main Authors | , , , , , , , |
---|---|
Format | Patent |
Language | English Korean |
Published |
22.10.2021
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | An apparatus and a method for embedding a sentence feature vector are disclosed. A sentence feature vector embedding apparatus in accordance with an embodiment of the present invention includes: a sentence acquisition unit configured to acquire a first sentence and a second sentence including one or more words; a vector extraction unit configured to independently input each of the first sentence and the second sentence into a feature extraction network and to extract a first feature vector corresponding to the first sentence and a second feature vector corresponding to the second sentence; and a vector compression unit configured to independently input each of the first feature vector and the second feature vector into a convolutional neural network-based (CNN) vector compression network and to compress each of the first feature vector and the second feature vector into a first compression vector and a second compression vector. The present invention can save time and resources consumed in training the feature extraction network.
문장 특징 벡터 임베딩 장치 및 방법이 개시된다. 일 실시예에 따른 문장 특징 벡터 임베딩 장치는, 하나 이상의 단어를 포함하는 제1 문장 및 제2 문장을 획득하는 문장 획득부; 상기 제1 문장 및 상기 제2 문장 각각을 특징 추출 네트워크에 독립적으로 입력하여 상기 제1 문장에 대응되는 제1 특징 벡터 및 상기 제2 문장에 대응되는 제2 특징 벡터를 추출하는 벡터 추출부; 및 상기 제1 특징 벡터 및 상기 제2 특징 벡터 각각을 컨볼루셔널 뉴럴 네트워크(CNN; Convolutional Neural Network) 기반의 벡터 압축 네트워크에 독립적으로 입력하여 상기 제1 특징 벡터 및 상기 제2 특징 벡터 각각을 제1 압축 벡터 및 제2 압축 벡터로 압축하는 벡터 압축부를 포함한다. |
---|---|
Bibliography: | Application Number: KR20200130918 |