BERT TEXTRANK BASED CORE SENTENCE EXTRACTION METHOD AND DEVICE USING BERT SENTENCE EMBEDDING VECTOR

摘要

The present invention relates to a method and apparatus for extracting a TextRank-based core sentence using a sentence embedding vector of an ERT, and the method for extracting a TextRank-based core sentence using a sentence embedding vector of a BERT according to an embodiment of the present invention is executed in a computing device A computer implemented method for extracting a core sentence to be used, comprising: a first step of dividing natural language data from which a core sentence is to be extracted into sentence units; a second step of adding a special classification token (CLS) before each of the divided sentences; a third step of converting each sentence to which the special classification token (CLS) is added into a sentence vector using a sentence vector transformation model; a fourth step of constructing a similarity matrix by calculating the similarity between sentences based on the sentence vectors; a fifth step of calculating the importance of each sentence by applying the similarity matrix to TextRank; and a sixth step of extracting a key sentence according to the calculated importance.