Dense video captioning aims to localize events within input videos and generate concise descriptive texts for each event. Advanced end-to-end methods require both tasks to share the same intermediate features that serve as event queries, thereby enabling the mutual promotion of two tasks. However, relying on shared queries limits the model’s ability to extract task-specific information, as event semantic perception and localization demand distinct perspectives on video understanding. To address this, we propose a decomposed dense video captioning framework that derives localization and captioning queries from event queries, enabling task-specific representations while maintaining inter-task collaboration. Considering the roles of different queries, we design a contrastive semantic optimization strategy that guides localization queries to focus on event-level visual features and captioning queries to align with textual semantics. Besides, only localization information is considered in existing methods for label assignment, failing to ensure the relevance of the selected queries to descriptions. We jointly consider localization and captioning losses to achieve a semantically balanced assignment process. Extensive experiments on the YouCook2 and ActivityNet Captions datasets demonstrate that our framework achieves state-of-the-art performance.

Zero-shot image captioning, which aims to generate image descriptions without relying on annotated data, has recently attracted increasing research interest. Pre-trained text-to-image generation models enable the creation of synthetic pairs solely from text data, while existing methods fall short in mitigating the discrepancy caused by the inability of synthetic images to fully capture the semantics of the textual input, resulting in unreliable cross-modal correspondences. To address this, we propose a retrieval-based framework that leverages only existing synthetic image-text pairs as its search corpus to systematically bridge the gap when using synthetic data for captioning. For the semantic gap between a synthetic image and its input text, our framework retrieves supplementary visual features from similar synthetic examples and integrates them to refine the image embedding. Then, it extracts image-related textual descriptions to mitigate the modality gap during decoding. Moreover, we introduce a plug-and-play visual semantic module that detects visual entities, further facilitating the construction of semantic correspondences between images and text. Experimental results on benchmark datasets demonstrate that our method obtains state-of-the-art results.

Adversarial attacks on deep neural networks keep raising security concerns in natural language processing research. Existing defenses focus on improving the robustness of the victim model in the training stage. However, they often neglect to proactively mitigate adversarial attacks during inference. Towards this overlooked aspect, we propose a defense framework that aims to mitigate attacks by confusing attackers and correcting adversarial contexts that are caused by malicious perturbations. Our framework comprises three components: (1) a synonym-based transformation to randomly corrupt adversarial contexts in the word level, (2) a developed BERT defender to correct abnormal contexts in the representation level, and (3) a simple detection method to filter out adversarial examples, any of which can be flexibly combined. Additionally, our framework helps improve the robustness of the victim model during training. Extensive experiments demonstrate the effectiveness of our framework in defending against word-level adversarial attacks.

Emotion cause analysis (ECA) aims to extract emotion clauses and find the corresponding cause of the emotion. Existing methods adopt fine-tuning paradigm to solve certain types of ECA tasks. These task-specific methods have a deficiency of universality. And the relations among multiple objectives in one task are not explicitly modeled. Moreover, the relative position information introduced in most existing methods may make the model suffer from dataset bias. To address the first two problems, this paper proposes a universal prompt tuning method to solve different ECA tasks in the unified framework. As for the third problem, this paper designs a directional constraint module and a sequential learning module to ease the bias. Considering the commonalities among different tasks, this paper proposes a cross-task training method to further explore the capability of the model. The experimental results show that our method achieves competitive performance on the ECA datasets.