Chinese keyword extraction method and device with multi-granularity semantic enhancement representation

• Main Authors: ZHANG SIMIN, GUO YU, YANG YUFEI, LI XUAN, LIU LIN, LU XIAO, WANG ZIXUAN, ZHOU XUANYU
• Format: Patent
• Language: Chinese; English
• Published: 26.07.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a multi-granularity semantic enhancement representation Chinese keyword extraction method and device, and the method comprises the steps: employing a Chinese keyword extraction network model of a field self-adaption module, a visual semantic enhancement representation module and a keyword sorting module to extract keywords from an input Chinese text; the field self-adaption module is used for extracting radical features from the Chinese text and fusing the radical features into a BERT-WWM model, so that field self-adaption text features of the Chinese text are extracted; the visual semantic enhancement representation module is used for extracting font features and carrying out visual semantic enhancement on text features based on a cross attention layer, and the keyword sorting module is used for predicting a tag sequence and labeling candidate keywords by utilizing a CRF model, and calculating word frequency-position weights of the candidate keywords to extract the keywords. The inven