The man-machine adaptability between soldiers and equipment affect the operational effectiveness of the whole individual soldier integrated system. In view of the complex semantic information and various equipment tendencies of individual soldier equipment evaluation, in order to extract key insights from soldiers’ comments on equipment, this paper constructs a Chinese multi-faceted and multi emotional (MME) dataset, and proposes an aspect-based sentiment analysis model. This model combines attention mechanism and context feature dynamic weighting method with global perceptual aspect embedding to enhance the BERT representation of aspect-based sentiment analysis. Then this study adds the offset correction to the ADAM algorithm, and finally maps the target to the correct context representation. The experimental results show that BERT’s dynamic weighting model has higher accuracy and score than MGAN, ATAE-LSTM, BERT-AEN, and BERT-Base models respectively in the classification of soldier equipment evaluation. It proves the effectiveness and feasibility of the data processing of soldier equipment evaluation.