Search Results (74)

Achieving optimal agricultural yields and promoting sustainable farming relies on accurate crop recommendations. However, the applicability of many current systems is limited by their considerable computational requirements and dependence on comprehensive datasets, especially in resource-limited contexts. This paper presents HOLISTIQ RS, a novel crop recommendation system explicitly designed for operation on low-specification hardware and in data-scarce regions. HOLISTIQ RS combines collaborative filtering with a Markov model to predict appropriate crop choices, drawing upon user profiles, regional agricultural data, and past crop performance. Results indicate that HOLISTIQ RS provides a significant increase in recommendation accuracy, achieving a MAP@5 of 0.31 and nDCG@5 of 0.41, outperforming standard collaborative filtering methods (the KNN achieved MAP@5 of 0.28 and nDCG@5 of 0.38, and the ANN achieved MAP@5 of 0.25 and nDCG@5 of 0.35). Significantly, the system also demonstrates enhanced recommendation diversity, achieving an Item Variety (IV@5) of 23%, which is absent in deterministic baselines. Significantly, the system is engineered for reduced energy consumption and can be deployed on low-cost hardware. This provides a feasible and adaptable method for encouraging informed decision-making and promoting sustainable agricultural practices in areas where resources are constrained, with an emphasis on lower energy usage. Full article

To address the limitations of existing knowledge graph-based recommendation algorithms, including insufficient utilization of semantic information and inadequate modeling of user behavior motivations, we propose SKGRec, a novel recommendation model that integrates knowledge graph and semantic features. The model constructs a semantic interaction graph (USIG) of user behaviors and employs a self-attention mechanism and a ranked optimization loss function to mine user interactions in fine-grained semantic associations. A relationship-aware aggregation module is designed to dynamically integrate higher-order relational features in the knowledge graph through the attention scoring function. In addition, a multi-hop relational path inference mechanism is introduced to capture long-distance dependencies to improve the depth of user interest modeling. Experiments on the Amazon-Book and Last-FM datasets show that SKGRec significantly outperforms several state-of-the-art recommendation algorithms on the Recall@20 and NDCG@20 metrics. Comparison experiments validate the effectiveness of semantic analysis of user behavior and multi-hop path inference, while cold-start experiments further confirm the robustness of the model in sparse-data scenarios. This study provides a new optimization approach for knowledge graph and semantic-driven recommendation systems, enabling more accurate capture of user preferences and alleviating the problem of noise interference. Full article

Knowledge graph-enhanced recommendation has gained increasing attention for its ability to provide structured semantic context. However, most existing approaches struggle with two critical challenges: the sparsity of long-tail relations in knowledge graphs and the lack of adaptability to users’ dynamic preferences. In this paper, we propose GLARA, a novel recommendation framework that combines semantic abstraction and behavioral adaptation through a two-stage modeling process. First, a Virtual Relational Knowledge Graph (VRKG) is constructed by clustering semantically similar relations into higher-level virtual groups, which alleviates relation sparsity and enhances generalization. Then, a global Local Weighted Smoothing (LWS) module and a local Graph Attention Network (GAT) are integrated to jointly refine item and user representations: LWS propagates information within each virtual relation subgraph to improve semantic consistency, while GAT dynamically adjusts neighbor importance based on recent interaction signals. Extensive experiments on Last.FM and MovieLens-1M demonstrate that GLARA outperforms state-of-the-art methods, achieving up to 5.8% improvements in NDCG@20, especially in long-tail and cold-start scenarios. Additionally, case studies confirm the model’s interpretability by tracing recommendation paths through clustered semantic relations. This work offers a flexible and interpretable solution for robust recommendation under sparse and dynamic conditions. Full article

Personalized recommendation for online learning courses stands as a critical research topic in educational technology, where algorithmic performance directly impacts learning efficiency and user experience. To address the limitations of existing studies in multimodal heterogeneous data fusion and high-order relationship modeling, this research proposes a Heterogeneous Hypergraph and Attention-based Online Course Recommendation (HHAOCR) algorithm. By constructing a heterogeneous hypergraph structure encompassing three entity types (students, instructors, and courses), we innovatively designed hypergraph convolution operators to achieve bidirectional vertex-hyperedge information aggregation, integrated with a dynamic attention mechanism to quantify important differences among entities. The method establishes computational frameworks for hyperedge-vertex coefficient matrices and inter-hyperedge attention scores, effectively capturing high-order nonlinear correlations within multimodal heterogeneous data, while employing temporal attention units to track the evolution of user preferences. Experimental results on the MOOCCube dataset demonstrate that the proposed algorithm achieves significant improvements in NDCG@15 and F1-Score@15 metrics compared to TP-GNN (enhanced by 0.0699 and 0.0907) and IRS-GCNet (enhanced by 0.0808 and 0.0999). This work provides a scalable solution for multisource heterogeneous data fusion and precise recommendation for online education platforms. Full article

Text-based image retrieval is one of the most common approaches for searching images acquired from vision sensors such as cameras. However, this method suffers from limitations in retrieval accuracy, particularly when the query contains limited information or involves previously unseen sentences. These challenges arise because keyword-based matching fails to adequately capture contextual and semantic meanings. To address these limitations, we propose a novel approach that transforms sentences and images into semantic graphs and scene graphs, enabling a quantitative comparison between them. Specifically, we utilize a graph neural network (GNN) to learn features of nodes and edges and generate graph embeddings, enabling image retrieval through natural language queries without relying on additional image metadata. We introduce a contrastive GNN-based framework that matches semantic graphs with scene graphs to retrieve semantically similar images. In addition, we incorporate a hard negative mining strategy, allowing the model to effectively learn from more challenging negative samples. The experimental results on the Visual Genome dataset show that the proposed method achieves a top nDCG@50 score of 0.745, improving retrieval performance by approximately 7.7 percentage points compared to random sampling with full graphs. This confirms that the model effectively retrieves semantically relevant images by structurally interpreting complex scenes. Full article

This study presents a comprehensive evaluation of embedding techniques and large language models (LLMs) for Information Retrieval (IR) and question answering (QA) across languages, focusing on English and Italian. We address a significant research gap by providing empirical evidence of model performance across linguistic boundaries. We evaluate 12 embedding models on diverse IR datasets, including Italian SQuAD and DICE, English SciFact, ArguAna, and NFCorpus. We assess four LLMs (GPT4o, LLama-3.1 8B, Mistral-Nemo, and Gemma-2b) for QA tasks within a retrieval-augmented generation (RAG) pipeline. We evaluate them on SQuAD, CovidQA, and NarrativeQA datasets, including cross-lingual scenarios. The results show multilingual models perform more competitively than language-specific ones. The embed-multilingual-v3.0 model achieves top nDCG@10 scores of 0.90 for English and 0.86 for Italian. In QA evaluation, Mistral-Nemo demonstrates superior answer relevance (0.91–1.0) while maintaining strong groundedness (0.64–0.78). Our analysis reveals three key findings: (1) multilingual embedding models effectively bridge performance gaps between English and Italian, though performance consistency decreases in specialized domains, (2) model size does not consistently predict performance, and (3) all evaluated QA systems exhibit a critical trade-off between answer relevance and factual groundedness. Our evaluation framework combines traditional metrics with innovative LLM-based assessment techniques. It establishes new benchmarks for multilingual language technologies while providing actionable insights for real-world IR and QA system deployment. Full article

As the e-commerce industry rapidly expands, the number of users and items continues to grow, making it increasingly difficult to capture users’ purchasing patterns. Sequential recommendation models have emerged to address this issue by predicting the next item that a user is likely to purchase based on their historical behavior. However, most previous studies have focused primarily on modeling item sequences using item IDs without leveraging rich item-level information. To address this limitation, we propose a sequential recommendation model called ITS-Rec that incorporates various types of textual item information, including item titles, descriptions, and online reviews. By integrating these components into item representations, the model captures both detailed item characteristics and signals related to purchasing motivation. ITS-Rec is built on a self-attention-based architecture that enables the model to effectively learn both the long- and short-term user preferences. Experiments were conducted using real-world Amazon.com data, and the proposed model was compared to several state-of-the-art sequential recommendation models. The results demonstrate that ITS-Rec significantly outperforms the baseline models in terms of Hit Ratio (HR) and Normalized Discounted Cumulative Gain (NDCG). Further analysis showed that online reviews contributed the most to performance gains among textual components. This study highlights the value of incorporating textual features into sequential recommendations and provides practical insights into enhancing recommendation performance through richer item representations. Full article

In the field of Information Retrieval (IR), user-submitted keyword queries often fail to accurately represent users’ true search intent. With the rapid advancement of artificial intelligence, particularly in natural language processing (NLP), query expansion (QE) based on large language models (LLMs) has emerged as a key strategy for improving retrieval effectiveness. However, such methods often introduce query topic drift, which negatively impacts retrieval accuracy and efficiency. To address this issue, this study proposes an LLM-based QE framework that incorporates a Gaussian kernel-enhanced semantic space for dense retrieval. Specifically, the model first employs LLMs to expand the semantic dimensions of the initial query, generating multiple query representations. Then, by introducing a Gaussian kernel semantic space, it captures deep semantic relationships among these query vectors, refining their semantic distribution to better represent the original query’s intent. Finally, the ColBERTv2 model is utilized to retrieve documents based on the enhanced query representations, enabling precise relevance assessment and improving retrieval performance. To validate the effectiveness of the proposed approach, extensive empirical evaluations were conducted on the MS MARCO passage ranking dataset. The model was systematically assessed using key metrics, including MAP, NDCG@10, MRR@10, and Recall@1000. Experimental results demonstrate that the proposed method outperforms existing approaches across multiple metrics, significantly improving retrieval precision while effectively mitigating query drift, offering a novel approach for building efficient QE mechanisms. Full article

In this paper, we address the challenge of dynamic evolution of user preferences and propose an attribute-sequence-based recommendation model to improve the accuracy and interpretability of recommendation systems. Traditional approaches usually rely on item sequences to model user behavior, but ignore the potential value of attributes shared among different items for preference characterization. To this end, this paper innovatively replaces items in user interaction sequences with attributes, constructs attribute sequences to capture fine-grained preference changes, and reinforces the prioritization of current interests by maintaining the latest state of attributes. Meanwhile, the item–attribute relationship is modeled using LightGCN and a variant of GAT, fusing multi-level features using gated attention mechanism, and introducing rotary encoding to enhance the flexibility of sequence modeling. Experiments on four real datasets (Beauty, Video Games, Men, and Fashion) showed that the model in this paper significantly outperformed the benchmark model in both NDCG@10 and Hit Ratio@10 metrics, with a highest improvement of 6.435% and 3.613%, respectively. The ablation experiments further validated the key role of attribute aggregation and sequence modeling in capturing user preference dynamics. This work provides a new concept for recommender systems that balances fine-grained preference evolution with efficient sequence modeling. Full article

The rapid growth in the global apparel market and the rise of online consumption underscore the necessity for intelligent clothing recommendation systems that balance visual compatibility with personalized preferences, particularly price sensitivity. Existing recommendation systems often neglect nuanced consumer price behaviors, limiting their ability to deliver truly personalized suggestions. To address this gap, we propose DeepFMP, a multimodal deep learning framework that integrates visual, textual, and price features through an enhanced DeepFM architecture. Leveraging the IQON3000 dataset, our model employs ResNet-50 and BERT for image and text feature extraction, alongside a comprehensive price feature module capturing individual, statistical, and category-specific price patterns. An attention mechanism optimizes multimodal fusion, enabling robust modeling of user preferences. Comparative experiments demonstrate that DeepFMP outperforms state-of-the-art baselines (LR, FM, Wide & Deep, GP-BPR, and DeepFM), achieving AUC improvements of 1.6–12.2% and NDCG@10 gains of up to 3.2%. Case analyses further reveal that DeepFMP effectively improves the recommendation accuracy, offering actionable insights for personalized marketing. This work advances multimodal recommendation systems by emphasizing price sensitivity as a pivotal factor, providing a scalable solution for enhancing user satisfaction and commercial efficacy in fashion e-commerce. Full article

This research aims to enhance e-recruitment systems using text summarization techniques and pretrained large language models (LLMs). A job recommender system is built with integrated text summarization. The text summarization techniques that are selected are BART, T5 (Text-to-Text Transfer Transformer), BERT, and Pegasus. Content-based recommendation is the model chosen to be implemented. The LinkedIn Job Postings dataset is used. The evaluation of the text summarization techniques is performed using ROUGE-1, ROUGE-2, and ROUGE-L. The results of this approach deduce that the recommendation does improve after text summarization. BERT outperforms other summarization techniques. Recommendation evaluations show that, for MRR, BERT performs 256.44% better, indicating relevant recommendations at the top more effectively. For RMSE, there is a 29.29% boost, indicating recommendations closer to the actual values. For MAP, a 106.46% enhancement is achieved, presenting the highest precision in recommendations. Lastly, for NDCG, there is an 83.94% increase, signifying that the most relevant recommendations are ranked higher. Full article

In contemporary society, individuals are inundated with a vast amount of redundant information, and recommendation systems have undoubtedly opened up new avenues for managing irrelevant data. Graph convolutional networks (GCNs) have demonstrated remarkable performance in the field of recommendation systems by iteratively performing node convolutions to capture information from neighboring nodes, thereby enhancing recommendation efficacy. However, most existing models fail to distinguish the importance of different nodes, which limits their performance. To address this issue, we propose the asGCL model. To mitigate the prevalent issue of popularity bias and to learn more uniform embedding representations, we have integrated a lightweight contrastive learning module into our model. Finally, extensive experiments conducted on four real-world datasets demonstrate the effectiveness of our model. Notably, on the Amazon-Books dataset, our asGCL model achieved improvements of 4.21% and 8.74% in recall@20 and NDCG@20, respectively. Full article

Conventional Applicant Tracking Systems (ATSs) encounter considerable constraints in accurately aligning resumes with job descriptions (JD), especially in handling unstructured data and intricate qualifications. We provide Resume2Vec, an innovative method that utilizes transformer-based deep learning models, including encoders (BERT, RoBERTa, and DistilBERT) and decoders (GPT, Gemini, and Llama), to create embeddings for resumes and job descriptions, employing cosine similarity for evaluation. Our methodology integrates quantitative analysis via embedding-based evaluation with qualitative human assessment across several professional fields. Experimental findings indicate that Resume2Vec outperformed conventional ATS systems, achieving enhancements of up to 15.85% in Normalized Discounted Cumulative Gain (nDCG) and 15.94% in Ranked Biased Overlap (RBO) scores, especially within the mechanical engineering and health and fitness domains. Although conventional the ATS exhibited slightly superior nDCG scores in operations management and software testing, Resume2Vec consistently displayed a more robust alignment with human preferences across the majority of domains, as indicated by the RBO metrics. This research demonstrates that Resume2Vec is a powerful and scalable method for matching resumes to job descriptions, effectively overcoming the shortcomings of traditional systems, while preserving a high alignment with human evaluation criteria. The results indicate considerable promise for transformer-based methodologies in enhancing recruiting technology, facilitating more efficient and precise candidate selection procedures. Full article

Data sparsity caused by limited interactions makes it challenging for recommendation to accurately capture user preferences. Contrastive learning effectively alleviates this issue by enriching embedding information through the learning of diverse contrastive views. The effectiveness of contrastive learning in uncovering users’ and items’ latent preferences largely depends on the construction of data augmentation strategies. Structure and feature perturbations are commonly used augmentation strategies in graph-based contrastive learning. Since graph structure augmentation is time consuming and can disrupt interaction information, this paper proposes a novel feature augmentation contrastive learning method. This approach leverages preference prototypes to guide user and item embeddings in acquiring augmented information. By generating refined prototypes for each user and item based on existing prototypes to better approximate true preferences, it effectively alleviates the over-smoothing issue within similar preferences. To balance feature augmentation, a prototype filtering network is employed to control the flow of prototype information, ensuring consistency among different embeddings. Compared with existing prototype-based methods, ProtoRec achieves maximum gains of up to 16.8% and 20.0% in recall@k and NDCG@k on the Yelp2018, Douban-Book, and Amazon-Book datasets. Full article

Recently, personalized news recommendation systems have been widely used, which can achieve personalized news recommendations based on people’s different preferences, optimize the reading experience, and alleviate the problem of information overload. Among them, session-based news recommendation has gradually become a research hotspot as it can recommend news without requiring users to log in or when their reading history is difficult to obtain. The key to session-based news recommendation is to use short-term interaction data to learn user preferences. Existing models often focus on mining news content information in sessions and do not fully utilize geolocation information related to news and sessions, and there is also a certain inconsistency between their training objective and model evaluation metric, leading to suboptimal model recommendation performance. In order to fully utilize geolocation information, this paper proposes a multi-level location-aware approach for session-based news recommendation (MLA4SNR). Firstly, a news-location heterogeneous graph is constructed, and a graph element-wise attention network is proposed to mine high-order relationships between news and location. Secondly, a session feature extraction network based on Transformer is proposed to extract session features. Then, a session-location heterogeneous graph is constructed, and a graph element-wise attention network is used to mine high-order relationships between sessions and locations. Finally, a loss function based on the NDCG is used to train the model. Experimental results on a real news dataset show that MLA4SNR outperforms the baselines significantly. Full article