In the cross-domain sentiment analysis, the labeled samples in the target domain are seriously insufficient, the distributions of features in different domains are very different, and the emotional polarities expressed by features in one domain differ a lot from the emotional polarities in another domain, all of these problems lead to low classification accuracy. To deal with the above problems, an aspect-level cross-domain sentiment analysis method based on capsule network was proposed. Firstly, the feature representations of text were obtained by BERT (Bidirectional Encoder Representation from Transformers) pre-training model. Secondly, for the fine-grained aspect-level sentiment features, Recurrent Neural Network (RNN) was used to fuse the context features and aspect features. Thirdly, capsule network and dynamic routing were used to distinguish overlapping features, and the sentiment classification model was constructed on the basis of capsule network. Finally, a small amount of data in the target domain was used to fine-tune the model to realize cross-domain transfer learning. The optimal F1 score of the proposed method is 95.7% on Chinese dataset and 91.8% on English dataset, which effectively solves the low accuracy problem of insufficient training samples.
To effectively improve the denoising effect of the original anisotropic diffusion model that used only the 4 neighborhood pixels information and ignored the diagonal neighborhood pixels information of the pixel to be repaired in the image denoising process, a image denoising algorithm using UK-flag shaped anisotropic diffusion model was proposed. This model not only made full use of the reference information of the 4 neighborhood pixels as in original algorithm, but also used another 4 diagonal neighborhood pixels information in the denoising process. Then the model using the 8 direction pixels information for image denoising was presented, and it was proved to be rational. The proposed algorithm, the original algorithm, and an improved similar algorithm were used to remove the noise from 4 images with noise. The experimental results show that the proposed algorithm has an average increase of 1.90dB and 1.43dB in Peak Signal-to-Noise Ratio (PSNR) value respectively, and an average increase of 0.175 and 0.1 in Mean Structure Similitary Index (MSSIM) value respectively, compared with the original algorithm and the improved similar algorithm, which concludes that the proposed algorithm is more suitable for image denoising. algorithm not only made full use of the reference information of the 4 neighborhood pixels as in original algorithm, but also another 4 diagonal neighborhood pixels information was used in the denoising process, and the algorithm was proved to be rationality. The experimental results showed that the proposed algorithm could increase the PSNR (peak signal-to-noise ratio) value 1.69db, and the MSSIM(mean structure similitary index) value 0.14, compared with the other similar algorithms in image denoising, which conclud that this proposed algorithm is more suitable for image denoising.
In order to deal with the problem that current digital rights expression models have less ability to describe dynamic semantics, a new model, DDRM(Dynamic Digical Rights Model), which can describe action state was presented. Based on first-order dynamic logic, a new symbol system of first-order dynamic logic, DrFDL(Digital rights Fist-order Dynamic Logic), was defined to describe digital rights conception DrFDL semantic structure which can reflect dynamic property of action was presented based on DDRM. In addition, a license syntax based on DDRM was provided for rights expression. Then DrFDL logic was used to express the formal semantics of the licenses produced from this syntax and the determinacy with validity of these licenses was explored at last.