Relation extraction (RE) aims to identify semantic relations between entities in unstructured text.Although recent work extends traditional RE to multimodal scenarios, most approaches still adopt classification-based paradigms with fused multimodal features, representing relations as discrete labels.This paradigm has two significant limitations: (1) it overlooks structural constraints like entity types and positional cues, and (2) it lacks semantic expressiveness for fine-grained relation understanding.We propose Retrieval Over Classification (ROC), a novel framework that reformulates multimodal RE as a retrieval task driven by relation semantics.ROC integrates entity type and positional information through a multimodal encoder, expands relation labels into natural language descriptions using a large language model, and aligns entity-relation pairs via semantic similarity-based contrastive learning.Experiments show that our method achieves state-of-the-art performance on the benchmark datasets MNRE and MORE and exhibits stronger robustness and interpretability.

In the field of Web knowledge base mining and Web services, the recognition service for named entity faces many challenges such as context complexity, semantic subtlety, and fuzzy entity boundaries, all of which require highly accurate and robust recognition service. The current services usually fail to reach those conditions. To address this issue, this paper proposes a recognition service for named entities for large Web knowledge bases, and the core contrition is the developed dual multilayer feature learning (D-MLFL) service, which combines projected gradient descent (PGD), adversarial learning, and a fused attention mechanism. Our service successfully addresses the challenges of complex context and subtle semantics faced by named entity recognition tasks. Our service integrates deep language models, recurrent neural networks, and conditional random fields, and clearly outperforms existing approaches in many sub-tasks, including in feature extraction, sequence modeling, and label decoding, especially in dealing with complex and diverse entity types and contextual relationships. We performed sufficient experiments, and the results show that our service significantly enhances the robustness against noise and abnormal Web data. The ability to extract entity features is improved, resulting in higher accuracy in identifying entities with fuzzy boundaries and complex semantics.

In representation learning for Web temporal knowledge bases, each node in Web knowledge bases carries a specific contextual meaning. Existing services often neglect the implicit semantic Web knowledge behind entities and relations, thus failing to effectively capture the knowledge representation of temporal Web knowledge bases. To address this issue, this paper develops an autocompletion service for temporal Web knowledge bases, which is based on multisource semantic feature learning and feature fusion. We construct a semantic model oriented toward external semantic Web repositories to supplement entity-relation descriptions, and our service leverages the pretrained language model BERT, effectively learning semantic knowledge features. Additionally, our service captures the textual features of quadruples using a recurrent neural network, constructs a historical sparse timestamp matrix, and generates a mask tensor, successfully obtaining the weights of potentially correct entities, and thereby capturing the historical features of quadruples. Furthermore, our service integrates complementary features from different modules through an attention mechanism. Experimental validation shows that our service outperforms existing approaches in terms of four evaluation metrics: mean reciprocal rank (MRR), Hits@1, Hits@3, and Hits@10. The results also exhibit that it improves the accuracy and performance for autocompletion service for temporal Web knowledge bases.

Abstract Sentiment classification is a key task in sentiment analysis, reviews mining, and other text mining applications. Various models have been proposed to build sentiment classifiers, but the classification performances of some existing methods are not good enough. Meanwhile, as a subproblem of sentiment classification, positive and unlabeled learning (PU learning) problem widely exists in real‐world cases, but it has not been given enough attention. In this article, we aim to solve the two problems in one framework. We first build a model for traditional sentiment classification based on adversarial learning, attention mechanism, and long short‐term memory (LSTM) network. We further propose an enhanced adversarial learning method to tackle PU learning problem. We conducted extensive experiments in three real‐world datasets. The experimental results demonstrate that our models outperform the compared methods in both traditional sentiment classification problem and PU learning problem. Furthermore, we study the effect of our models on word embedding. Finally, we report and discuss the sensitivity of our models to parameters.