Temporal Knowledge Graph Question Answering (TKGQA) task aims to find an entity or timestamp from a temporal knowledge graph to answer temporal reasoning questions. However, most existing models fail to capture the implicit temporal information in the questions, resulting in weak performance when handling complex temporal reasoning tasks. To address this issue, this paper proposes a novel TKGQA model called GATQR, which integrates graph attention mechanism. The model utilizes a pre-trained temporal knowledge base in the form of quadruples and introduces Graph Attention Network (GAT) to effectively capture the implicit temporal information in the questions. By integrating with relation representations trained by the RoBERTa, it further enhances the temporal relationship representation in the queries. Finally, this representation is combined with the pre-trained TKG embeddings to predict the entity or timestamp with the highest score as the answer. Experimental results on the largest benchmark dataset CronQuestion demonstrate that compared to baseline models such as CronKGQA, EntityQR, and TempoQR-Soft, the GATQR achieves significant improvements in Hits@l results for handling complex and temporal question types, with increases of 35% and 13%, 18% and 9%, and 9% and 3%, respectively. These results validate the effectiveness and superiority of the GATQR model in capturing implicit temporal information and enhancing complex reasoning capabilities.

Vision Language Pre-training (VLP) has made significant progress in the field of universal multimodality in recent years. Universal multimodal datasets (such as MSCOCO, Flickr30k, etc.) have become one of the standards for evaluating VLP models which rely on images and corresponding captions for representation modeling. However, in a public opinion event, besides image captions, it also includes other texts such as content texts and comments, which may have a positive impact on image-text representations of public opinion. In this paper, we propose the method to explore the positive effect of content texts on initial multimodal representations. We name our model CLIP-PubOp, which is based on CLIP, a famous VLP model using contrastive learning. We add a linear fusion layer before the fusion of image and text representations, which fuses event content text representations and caption representations in a certain proportion to obtain enhanced text representations. On this basis, multimodal representations can be obtained by fusing enhanced text representations with image representations. We crawl through four mainstream categories of public opinion events online as our datasets and conduct experiments on both Chinese and English version of datasets. The experimental results show that content texts of public opinion events have a significant positive effect on multimodal representations, with an average accuracy improvement of about 2%-10% in image-text retrieval tasks.