Search

The goal of multimodal named entity recognition (MNER) is to detect entity spans in given image–text pairs and classify them into corresponding entity types. Despite the success of existing works that leverage cross-modal attention mechanisms to integrate textual and visual representations, we observe three key issues. Firstly, models are prone to misguidance when fusing unrelated text and images. Secondly, most existing visual features are not enhanced or filtered. Finally, due to the independent encoding strategies employed for text and images, a noticeable semantic gap exists between them. To address these challenges, we propose a framework called visual clue guidance and consistency matching (GMF). To tackle the first issue, we introduce a visual clue guidance (VCG) module designed to hierarchically extract visual information from multiple scales. This information is utilized as an injectable visual clue guidance sequence to steer text representations for error-insensitive prediction decisions. Furthermore, by incorporating a cross-scale attention (CSA) module, we successfully mitigate interference across scales, enhancing the image’s capability to capture details. To address the third issue of semantic disparity between text and images, we employ a consistency matching (CM) module based on the idea of multimodal contrastive learning, facilitating the collaborative learning of multimodal data. To validate the effectiveness of our proposed framework, we conducted comprehensive experimental studies, including extensive comparative experiments, ablation studies, and case studies, on two widely used benchmark datasets, demonstrating the efficacy of the framework.

Chinese spelling correction (CSC) constitutes a pivotal and enduring goal in natural language processing, serving as a foundational element for various language-related tasks by detecting and rectifying spelling errors in textual content. Numerous methods for Chinese spelling correction leverage multimodal information, including character, character sound, and character shape, to establish connections between incorrect and correct characters. Research indicates that a majority of spelling errors stem from pinyin similarity, with character similarity accounting for half of the errors. Consequently, effectively modeling character pinyin and character relationships emerges as a key challenge in the CSC task. In this study, we propose enhancing the CSC task by introducing the pinyin character prediction task. We employ an adaptive weighting method in the pinyin character prediction task to address predictions in a more granular manner, achieving a balance between the two prediction tasks. The proposed model, SPMSpell, utilizes ChineseBERT as an encoder to capture multimodal feature information simultaneously. It incorporates three parallel decoders for character prediction, pinyin prediction, and self-distillation modules. To mitigate potential overfitting concerning pinyin, a self-distillation method is introduced to prioritize character information in predictions. Extensive experiments conducted on three SIGHAN benchmark tests showcase that the model introduced in this paper attains a superior level of performance. This substantiates the correctness and superiority of the adaptive weighted pinyin character prediction task and underscores the effectiveness of the self-distillation module.

In recent years, with the advancement of natural language processing techniques and the release of models like ChatGPT, how language models understand questions has become a hot topic. In handling complex logical reasoning with pre-trained models, its performance still has room for improvement. Inspired by DAGN, we propose an improved DaGATN (Discourse-apperceptive Graph Attention Networks) model. By constructing a discourse information graph to learn logical clues in the text, we decompose the context, question, and answer into elementary discourse units (EDUs) and connect them with discourse relations to construct a relation graph. The text features are learned through a discourse graph attention network and applied to downstream multiple-choice tasks. Our method was evaluated on the ReClor dataset and achieved an accuracy of 74.3%, surpassing the best-known performance methods utilizing deberta-xlarge-level pre-trained models, and also performed better than ChatGPT (Zero-Shot).

Open-domain event extraction is a fundamental task that aims to extract non-predefined types of events from news clusters. Some researchers have noticed that its performance can be enhanced by improving dependency relationships. Recently, graphical convolutional networks (GCNs) have been widely used to integrate dependency syntactic information into neural networks. However, they usually introduce noise and deteriorate the generalization. To tackle this issue, we propose using Bi-LSTM to obtain semantic representations of BERT intermediate layer features and infuse the dependent syntactic information. Compared to current methods, Bi-LSTM is more robust and has less dependency on word vectors and artificial features. Experiments on public datasets show that our approach is effective for open-domain event extraction tasks.

The extant event detection models, which rely on dependency parsing, have exhibited commendable efficacy. However, for some long sentences with more words, the results of dependency parsing are more complex, because each word corresponds to a directed edge with a dependency parsing label. These edges do not all provide guidance for the event detection model, and the accuracy of dependency parsing tools decreases with the increase in sentence length, resulting in error propagation. To solve these problems, we developed an event detection model that uses a self-constructed dependency and graph convolution network. First, we statistically analyzed the ACE2005 corpus to prune the dependency parsing tree, and combined the named entity features in the sentence to generate an undirected graph. Second, we implemented an enhanced graph convolution network using the multi-head attention mechanism to understand the representation of nodes in the graph. Finally, a gating mechanism combined the semantic and structural dependency information of the sentence, enabling us to accomplish the event detection task. A series of experiments conducted on the ACE2005 corpus demonstrates that the proposed method enhances the performance of the event detection model.

Named entity recognition aims to extract entities with specific meaning from unstructured text. Currently, deep learning methods have been widely used for this task and have achieved remarkable results, but it is often difficult to achieve better results with less labeled data. To address this problem, this paper proposes a method for cross-lingual entity recognition based on an attention mechanism and adversarial training, using resource-rich language annotation data to migrate to low-resource languages for named entity recognition tasks and outputting changing semantic vectors through the attention mechanism to effectively solve the long-sequence semantic dilution problem. To verify the effectiveness of the proposed method, the method in this paper is applied to the English–Chinese cross-lingual named entity recognition task based on the WeiboNER data set and the People-Daily2004 data set. The obtained F1 value of the optimal model is 53.22% (a 6.29% improvement compared to the baseline). The experimental results show that the cross-lingual adversarial named entity recognition method proposed in this paper can significantly improve the results of named entity recognition in low resource languages.

Previous work has demonstrated that end-to-end neural sequence models work well for document-level event role filler extraction. However, the end-to-end neural network model suffers from the problem of not being able to utilize global information, resulting in incomplete extraction of document-level event arguments. This is because the inputs to BiLSTM are all single-word vectors with no input of contextual information. This phenomenon is particularly pronounced at the document level. To address this problem, we propose key-value memory networks to enhance document-level contextual information, and the overall model is represented at two levels: the sentence-level and document-level. At the sentence-level, we use BiLSTM to obtain key sentence information. At the document-level, we use a key-value memory network to enhance document-level representations by recording information about those words in articles that are sensitive to contextual similarity. We fuse two levels of contextual information by means of a fusion formula. We perform various experimental validations on the MUC-4 dataset, and the results show that the model using key-value memory networks works better than the other models.