APPARATUS AND METHOD FOR EXTRACTING SIGNIFICANT TERMS WITHIN A DOCUMENT

摘要

A device and a method for extracting significant terms from a document are provided to extract the significant terms of the document by using PCA(Principal Component Analysis) and SVD(Single Value Decomposition) without using information tools like a large capacity corpus or thesaurus. A matrix generator(11) generates a multidimensional space using each terms of the document as coordinate axes and generates a sentence-term matrix corresponding to the multidimensional space. An SVD part(12) removes noise by performing the SVD for the generated sentence-term matrix. A PCA part(13) performs the PCA for the noise removed sentence-term matrix, represents each principal component selected as a result with an eigenvector coefficient, and quantizes each principal component to an eigenvalue. A significant term extractor extracts the significant term from the principal components by considering an accumulation ratio and a principal component loading coefficient of the eigenvalue corresponding to each quantized principal component.