Aspect-based sentiment analysis (ABSA) predicts the sentiment polarity of specific aspects within a sentence, providing a finer-grained analysis than traditional text classification. Most existing methods use graph neural networks (GNNs) or multitask learning to model word relationships, while the direct application of transformers to ABSA remains underexplored. We propose a Unified Syntax-Guided Transformer (USGT) for analyzing the aspect-based sentiment of a sentence. USGT computes syntactic distances via dependency trees and models node importance by an inverse-distance weighting function, integrating this information into the transformer’s attention mechanism. In a multitask framework, USGT performs aspect extraction and sentiment classification jointly, using shared syntax-guided features. A Local-Global Representation Fusion (LGRF) layer combines local aspect-based pooling with global sentence representations to improve contextual modelling.Experiments on multiple ABSA benchmarks demonstrate the effectiveness of USGT.

Deep neural networks are widely used in image classification tasks, but their internal decision-making mechanisms are often difficult to explain. While various algorithms have been developed to visualize these mechanisms, many of them produce coarse, noisy results that are not always convincing. To address this issue, we propose a method for enhancing saliency maps produced by saliency methods. Our method uses a fixed-size sliding window to upsample local regions of the input image and feed them into the selected visualization algorithm to generate class-specific saliency maps and probability scores. We then downsample the resulting saliency maps and multiply them by the probability scores to obtain maps with greater detail. We evaluate our method using different saliency methods and network architectures, and demonstrate its effectiveness through both quantitative metrics and intuitive evaluation. Our results show that our method significantly improves the performance of these saliency methods, providing a more valid and reliable means of visualizing the decision mechanisms of deep neural networks. Code is available at https://github.com/LuoLogic/Enhuncement-saliency.

The visualization of deep learning models has been widely studied as an effective means of exploring the decision-making processes within these models. However, current visualization methods suffer from several limitations, such as low resolution and poor visualization of multiple occurrences of the same class. In this paper, we propose a novel visualization technique called MSG-CAM, which is an improvement on the existing Group-CAM method. Our method uses the feature maps and gradients of the last layer of the convolutional neural network to create masks through multi-scale enlargement of the original input image and fusion of the resulting feature maps and gradients. Through both qualitative and quantitative analysis, we have demonstrated that the saliency maps generated by our method are more reasonable and accurately reflect the internal decision-making processes of the neural network.

Most existing methods on affective video content analysis are dedicated to first obtain visual and audio representations, and then fuse them into a compact representation. However, these fusion techniques are often challenged by the modality gaps between heterogeneous modalities. In this paper, we aim to learn effective modality representations to aid the process of fusion. We propose a novel framework named Private-Shared Subspaces Learning(P2SL), which projects each modality into two distinct subspaces. The first subspace is shared by all modalities, where the representations across modalities learn their commonalities and reduce the modality gap. The second subspace is private to each modality and captures their characteristic features. Specifically, we introduce inter-modal similarity loss and intra-modal loss to obtain the robust features. In order to capture cues in different grains, global and local features are extracted from both visual and audio signals. Experiments show that our proposed approach outperforms the state-of-the-art methods and demonstrate the effectiveness of our approach.

In recent times, the proliferation of multimedia devices and reduced costs of data storage have enabled people to easily record and collect a large number of home videos; furthermore, this collection is growing with time. With the popularity of participatory media such as YouTube and facebook, problems are encountered when people intend to share their home videos with others. The first problem is that different people might be interested in different video content. Given the numbers of home videos, it is a time-consuming and hard task to manually select proper content for people with different interests. Secondly, as short videos are becoming more and more popular in media sharing applications, people need to manually cut and edit home videos which is again a tedious task. In this paper, we propose a method that employs affective analysis to automatically create video presentations from home videos. Our novel method adaptively creates presentations based on three properties: emotional tone, local main character and global main character. A novel sparsity-based affective labeling method is proposed to identify the emotional content of the videos. The local and global main characters are determined by applying face recognition in each shot. To demonstrate the proposed method, three kinds of presentations are created for family, acquaintance and outsider. Experimental results show that our method is very effective in video sharing and the users are satisfied with the videos generated by our method.