A FRAMEWORK FOR SEMANTIC DOCUMENT SIMILARITY DETECTION BASED ON AUTOMATIC TEXT SUMMARIZATION AND TRANSFORMER EMBEDDINGS

The rapid growth of digital document collections has increased the need for efficient methods for identifying semantically similar documents. Traditional similarity measures often rely on lexical overlap and fail to capture deeper semantic relationships. This paper proposes a framework for semantic document similarity detection based on automatic text summarization and transformer embeddings. First, source documents are summarized to extract their most informative content while reducing noise and redundancy. The generated summaries are then encoded using transformer-based embedding models to obtain semantic representations. Document similarity is computed in the embedding space using cosine similarity. The proposed approach improves retrieval efficiency and semantic matching accuracy for large-scale document collections

Hali tarjima qilinmagan