Multi-level fusion aspect category sentiment analysis method based on self-attention

• Main Authors: TIAN DONG, ZHANG HUI, FENG JIANYING, TOKIYOSHI, MIAO ZIWEI
• Format: Patent
• Language: Chinese; English
• Published: 04.07.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a multi-level fusion aspect category sentiment analysis method based on self-attention, and belongs to the field of sentiment word analysis by applying a self-attention mechanism. The method mainly comprises the following steps: acquiring a sample statement and a plurality of groups of two-tuple tag information associated with the sample statement; obtaining sequence information and local important information of the input text by combining Bi-LSTM and a convolution attention mechanism; acquiring position information of aspect category indication words; according to the method, aspect information is learned again by a scene-containing self-attention mechanism, then a multi-level fusion module is introduced to perform fusion learning on interaction among multi-level context features, aspect category sentiment analysis is realized, aspect representation in different environments is concerned differently, and better classification precision can be achieved in multiple fields. The inventio