[Objective] To perceive the popularity preferences of users and their friends towards items, more accurate recommendation service can be achieved. [Methods] This paper proposes an item popularity calculation method that integrates contribution and influence. Attention mechanism and recurrent neural network are used to capture user popularity preference representation, and convolutional neural network and graph attention mechanism are also used to obtain friends’ long-term and short-term popularity preferences. [Results] Comparative experiments are conducted using the Douban, Delicious and Yelp datasets, and the evaluation metrics of this method are superior to the suboptimal model DGRec. The highest value of Recall@20 increases by 13.03%, and the highest increase of NDCG is 11.69%. Compared to traditional calculation methods, the proposed popularity calculation method achieves the highest increase in Recall@20 by 11.53%, and the highest increase in NDCG by 10.29%. [Limitations] This method still needs to improve performance when dealing with short sequences. [Conclusions] This method adds user popularity preference representation and user social popularity preference representation, enhances the ability to express the weight of each interaction, and can effectively recommend more long-tail items.
[Objective] This paper explores the impact of product domain knowledge and dynamic image-text interactions on the helpfulness of reviews, aiming to improve the performance of multimodal review helpfulness identification. [Methods] We introduced a domain knowledge-enhanced method for multimodal review helpfulness detection. First, we identified domain keywords based on the implicit topic information in the reviews and obtained a domain knowledge feature representation using a topic attention mechanism. Then, we developed a knowledge-enhanced dynamic text-image interaction module. This module utilized a knowledge-enhanced intra-modality self-attention mechanism that seamlessly integrated domain knowledge into both textual and visual representations. Finally, we used a knowledge-enhanced inter-modal co-attention mechanism to obtain the feature representation resulting from dynamic interaction between the knowledge-enhanced text and image. [Results] Our model’s F1-score on the Amazon dataset reached 89.57%, which was 0.9 percentage points higher than the best baseline model. [Limitations] This paper only examined the model on English datasets. Its performance on Chinese datasets needs further investigation. [Conclusions] The domain knowledge-enhanced model effectively improves the performance of identifying review helpfulness. It also extracts key information from images and texts, enhancing the model’s interpretability.
[Objective] To address the insufficient fusion of modalities in multimodal rumour detection, we propose MNEF, a model designed to enhance modal fusion and improve detection accuracy. [Methods] The extracted features of the image processed in the frequency domain serve asa complementary modality. A feature fusion approach is then employed to integrate the feature vectors of text, image, and frequency domain modalities, enabling the model to capture deeper semantic relationships among them. [Results] Compared to the optimal baseline models, the MNEF model achieved accuracyimprovements of 3.02% and 0.81%, respectively. Furthermore, ablation experiments revealed that the MNEF model outperformed the four ablation models in accuracy by 1.51%, 4.68%, 5.07% and 4.61%. [Limitations] With the addition of the frequency domain processing branch, the model's overall complexity and computational cost have increased. However, it is limited to processing images and text, failing to account for rumours in other modalities. [Conclusions] By incorporating frequency domain processing and feature fusion for modality enhancement, the model can capture deeper semantic relationships among modalities.
[Objective] This study proposes a research framework for risk analysis of telecom fraud based on large language models (LLMs) and event fusion to reveal the process of telecom fraud and identify key risk factors. [Methods] We constructed a two-stage hierarchical prompt instruction specific to the telecom fraud domain and extracted risk events and their arguments from fraud cases. The framework integrates semantic dependency analysis with template-matching techniques to obtain the fraud event chains. Considering the diversity in event descriptions, we employed the BERTopic model for sentence vector representation and utilized a clustering algorithm for event fusion. [Results] Our method achieved F1-scores of 67.41% for event extraction and 73.12% for argument extraction in telecom fraud case analysis. Event clustering identified 10 categories of thematic risk events, with “disclosing information” as the highest-risk behavior. [Limitations] The coarse granularity of police report data limits the framework’s early warning capabilities. [Conclusions] The proposed approach, combining LLMs with event fusion clustering, enables the automatic construction of fraud event evolution chains, facilitates risk analysis, and supports the early warning and deterrence of telecom frauds.
[Objective] To address the issues in multimodal sentiment analysis, such as insufficient multimodal feature extraction, semantic differences between modalities, and lack of interaction, we propose a temporal multimodal sentiment analysis model that integrates multi-level attention and sentiment scale vectors. [Methods] Firstly, we introduced a scalar Long Short-Term Memory network with a multi-head attention mechanism to construct a deep temporal feature modeling network for extracting rich contextual temporal features from text, audio, and visual modalities. Secondly, we employed the text-guided dual-layer cross-modal attention mechanism and the improved self-attention mechanism to facilitate the deep information exchange across modalities, thereby generating two sentiment scale vectors for sentiment intensity and polarity. Finally, the L1 norm of the sentiment intensity vector was multiplied by the normalized sentiment polarity vector to obtain a comprehensive representation of sentiment strength and polarity, thereby enabling accurate sentiment prediction. [Results] Experiments on the CMU-MOSI dataset show that the proposed model achieves good results in both comparative and ablation experiments, outperforming the next-best model by 1.2 and 2.3 percentage points on the Acc7 and Corr metrics, respectively. On the CMU-MOSEI dataset, the proposed model surpasses baseline models across all evaluation metrics, achieving 86.0% in Acc2 and 86.1% in F1 score. [Limitations] Sentiment expression is highly context-dependent, and the sources of sentiment cues may vary across different scenarios. The proposed model may perform poorly when textual information is insufficient. [Conclusions] The proposed model effectively extracts contextual temporal features from various modalities and leverages the rich emotional information in the text modality for deep inter-modal interaction, thereby enhancing the accuracy of sentiment prediction.
[Objective] This study proposes a method for discovering knowledge combinations that are sensitive to the relationships among low- and medium-frequency knowledge elements (KE). It integrates the semantic features of the value and importance of multi-source knowledge units by examining both scientific research outputs and social media content. [Methods] First, we designed and calculated semantic features representing the KE value and importance from the perspectives of scientific research and social media. Then, we constructed a heat conduction model to integrate multiple semantics and uncover latent associations between KEs and documents. Finally, based on the newly constructed KE-document network, we calculated the weighted Jaccard coefficients between KEs to enable knowledge combination discovery. [Results] We conducted empirical validation using CSSCI literature in Information Science and Baidu Baike. Compared with methods without feature integration, the proposed method showed performance improvements of 0.300, 0.230, 0.184, 0.183, and 0.278 under the P@50, P@100, P@500, P@1000, and P@2000 evaluation metrics, respectively. Some knowledge combinations not verified in the literature, such as “government information resources-industry think tank alliance”, “Weibo comments-metaphor recognition”, and “social impact theory-stochastic resonance”, also demonstrate high combination potential and interpretability. [Limitations] The discovered knowledge combinations have not been further evaluated or analyzed; fine-grained semantic relationships between knowledge units remain an unresolved challenge. [Conclusions] The proposed method is reliable and effective, offering valuable reference for future research. The discovered combinations can provide insights into academic innovation and disciplinary development.
[Objective] This study investigates the patterns and characteristics of key actors in international scientific collaboration through the lens of research content. [Methods] Utilizing large language models and scientific memes to gauge content homogeneity, cooperation patterns are categorized based on homogeneity and frequency, with a quantitative analysis of actor characteristics in terms of partners and content. [Results] The empirical study was carried out in the field of brain-computer interface. Core countries predominantly exhibit homogeneous and heterogeneous cooperation models, with free cooperation primarily involving exploratory ventures in emerging fields and dependent small country collaborations. Most nations lean towards international cooperation over domestic independence, with a high degree of partner homogeneity in both multilateral and bilateral alliances for leading and dependent small countries. Nations such as China and the UK favor international cooperation grounded in their research content, in contrast to the US and Germany, which show a preference for collaborative research into new topics. [Limitations] This study focuses solely on content homogeneity, without incorporating other dimensions of homogeneity into the analytical framework, resulting in a lack of comprehensiveness. [Conclusions] This research enriches the analytical tools for international collaboration patterns, offering a new lens for understanding global scientific partnerships.
[Objective] This study examines the Management Discussion and Analysis (MD&A) section in annual reports of listed companies to reveal its informational value for predicting financial distress, aiming to enhance the accuracy and practicality of financial risk assessments. [Methods] We employed a Hierarchical Attention Network (HAN) to build a semantic model of the annual reports. Then, we extracted semantic vectors from the report’s MD&A section with the model and combined them with the report’s tone to predict financial distress. Finally, we utilized heatmaps to highlight key contents indicating financial risks. [Results] The semantic model achieved strong predictive accuracy (AUC=0.895), with MD&A’s opening/closing sentences and tonal features playing a crucial role. Key information was mainly concentrated in five dimensions: policy, operations, performance, governance, and risk. [Limitations] The study analyzed only annual report texts without incorporating financial indicators or other valuable data. Future research can integrate annual reports with other related information to enhance prediction accuracy and comprehensiveness. [Conclusions] This study confirms the importance of semantic features in MD&A texts for financial forecasting using the HAN model. It offers scientific support for investor decision-making and improves corporate information disclosure.
[Objective] This study explores the impact of the voice characteristics of AI medical voice assistants on the perceived credibility of the older adults, mainly based on the Computer Are Social Actors (CASA) paradigm and the stereotype model. [Methods] This study conducted a 3 (voice gender: female/male/non-binary) ×2 (communication style: expert/partner) between-subjects experiment to explore the impact of voice gender and communication style of AI medical voice assistants on the perceived credibility and intentions to use among older adults. Additionally, the study sought to elucidate the mechanism of action on the stereotype dimensions of perceived warmth and perceived professionalism. [Results] The results indicate that older adults perceive male expert-type and female partner-type AI medical voice assistants as more credible. Communication style influenced their credibility perception of voice gender through perceived professionalism, and this perceived credibility positively predicted their behavioral intention to use such assistants. [Limitations] As this study was conducted within the context of China’s smart healthcare system development, the generalizability of the findings warrants further validation. [Conclusions] The congruence between vocal characteristics and gender-role stereotypes enhanced older adults’ perceived credibility. AI medical voice assistant design should account for the interplay of multiple vocal factors and contextual suitability.
[Objective] This work aims to address the challenge of scarce supervised data in classical Chinese entity extraction by leveraging knowledge distillation techniques to inject knowledge from unsupervised external sources into a student model. [Methods] A large language model is utilized as a generative knowledge teacher model to perform knowledge distillation on unsupervised corpora. Additionally, a dictionary knowledge teacher model is built using supervised data from the ZuoZhuan and GuNer datasets. The knowledge distilled from both teachers is integrated to compile a semi-supervised dataset for classical Chinese entity extraction. The task is then reformulated as a sequence-to-sequence problem, and pre-trained models such as mT5 and UIE are fine-tuned on this dataset. [Results] On the ZuoZhuan and GuNer datasets, the proposed method achieves F1-Score of 89.15% and 95.47%, respectively, outperforming the baseline models SikuBERT and SikuRoBERTa, which were incrementally fine-tuned on classical Chinese corpora, by 8.15% and 9.27% in F1-Score. [Limitations] The method does not incorporate additional entity type information, and the quality of data pre-retrieved by the LLMs may affectt extraction results. [Conclusions] In low-resource settings, the proposed approach effectively distills the knowledge advantages of pre-trained large language models and dictionary resources into the student entity extraction model, significantly improving the performance on classical Chinese entity extraction tasks.
[Objective] This study addresses the issues of low fine-tuning efficiency and poor extraction performance in the tourism domain caused by dispersed knowledge and limited annotated data. It explores entity-relation extraction methods under few-shot learning scenarios. [Methods] Based on the large model GLM, we applied prompt learning in the tourism domain to encode input text. Then, we utilized a global pointer network to predict potential relations and identify entities with specific relations, thereby extracting relation triplets. [Results] We conducted experiments on the self-constructed tourism dataset and the Baidu DuIE dataset. The F1 values of our model were 90.51% and 89.45%, which were 2.37% and 0.16% higher than the traditional relationship extraction models. [Limitations] The prompt learning approach was only applied in the tourism domain and with specific encoders, indicating room for expansion in application scenarios. [Conclusions] The proposed method improves entity-relation joint extraction in tourism texts. Prompt learning and large language model encoders can alleviate the performance issues associated with model training, effectively improving the accuracy of entity-relation extraction.
[Objective] This paper addresses the issues of overlapping entities and complex relationships in Traditional Chinese Medicine (TCM) patent texts. It improves the extraction accuracy of entity relationship, such as TCM components, pharmacological effects, and advantages. [Methods] We proposed a joint entity-relationship extraction model for TCM patent texts named TPSCRE. This model combines a Part-of-Speech (POS) tagging network with a CDIL-CNN enhanced architecture to enhance semantic understanding of TCM patent texts. We used a dual cross-attention mechanism to generate diverse word representations, facilitating interaction and complementarity between entities and relations. We also employed an adversarial learning strategy to improve model robustness and generalization ability against potentially mislabeled data. Finally, we constructed a subject-object correspondence matrix to filter out correct entity-relation triplets from TCM patents. [Results] We conducted comparative and ablation experiments on a self-constructed TCM patent dataset,and the proposed TPSCRE model achieved optimal performance, with F1 scores of 94.71% for TCM entity recognition and 87.56% for relationship extraction. [Limitations] The model has high complexity and computational cost. The scale of existing datasets constrains its evaluation metrics. [Conclusions] The TPSCRE model effectively captures complex entity relationships in TCM texts, achieving significant performance advantages in the joint extraction task of entity relations from the TCM patent texts.
[Objective] This paper proposes a prompt template-guided document-level event extraction model (DEEM-PT), aiming to address the issues of argument dispersion and multiple events in document-level financial event extraction. [Methods] The DEEM-PT model designed prompt templates based on financial event types and used graph neural networks with pseudo-event proxy nodes to enhance the associations among arguments, sentences, and events. The model also reinforced information interaction across multiple events. Additionally, we employed a multi-head attention mechanism to integrate features such as prompt templates, arguments, and events. [Results] We examined the DEEM-PT model on the ChFinAnn dataset, and it performed excellently across various financial events, with precision, recall, and F1 values reaching 85.2%, 81.5%, and 83.3%, respectively. [Limitations] The training of DEEM-PT relies on a financial event dataset. Therefore, the design of prompt templates depends on domain knowledge and expert input. [Conclusions] Introducing event prompt templates and enhancing information interaction in graph neural networks can effectively improve the model’s classification performance for event types and arguments.
[Objective] This paper constructs a retrieval-augmented question-answering (QA) system for Traditional Chinese Medicine (TCM) standards, aiming to provide efficient standard knowledge services and promote the research and application of TCM standardization. [Methods] By comparing the performance of large language models such as BaiChuan, Gemma, and Qwen, we chose GPT-3.5 as the base model. Then, we combined data optimization and retrieval-augmented generation to develop a QA system with semantic analysis, contextual association, and answer-generation capabilities. [Results] On a TCM literature-based question generation dataset, the new system achieved answer relevance precision, recall, and F1 scores of 0.879, 0.839 and 0.857, respectively, as well as contextual relevance scores of 0.838, 0.869, and 0.853. On a TCM standards QA dataset, the system achieved answer relevance scores of 0.871, 0.836 and 0.853, all outperforming baseline models. [Limitations] The system’s intent recognition accuracy still requires further improvement. The scale and granularity of the TCM standards knowledge base need to be expanded and refined. [Conclusions] In response to the practical needs of TCM knowledge services, this study developed a retrieval-augmented QA system for TCM standards. The system can effectively answer various questions related to clinical guidelines, herbal medicine standards, and information standards, covering topics such as treatment principles, syndrome classification, therapeutic methods, and technical specifications, demonstrating its strong practicality and feasibility.
