Search Results (44)

Traditional retrieval-augmented generation (RAG) methods struggle with hierarchical documents, often causing semantic fragmentation, structural loss, and inefficient retrieval due to fixed strategies. To address these challenges, this paper proposes HiSem-RAG, a hierarchical semantic-driven RAG method. It comprises three key modules: (1) hierarchical semantic indexing, which preserves boundaries and relationships between sections and paragraphs to reconstruct document context; (2) a bidirectional semantic enhancement mechanism that incorporates titles and summaries to facilitate two-way information flow; and (3) a distribution-aware adaptive threshold strategy that dynamically adjusts retrieval scope based on similarity distributions, balancing accuracy with computational efficiency. On the domain-specific EleQA dataset, HiSem-RAG achieves 82.00% accuracy, outperforming HyDE and RAPTOR by 5.04% and 3.98%, respectively, with reduced computational costs. On the LongQA dataset, it attains a ROUGE-L score of 0.599 and a BERT_F1 score of 0.839. Ablation studies confirm the complementarity of these modules, particularly in long-document scenarios. Full article

To address the limitations of Retrieval-Augmented Generation (RAG) systems in handling long policy documents, mitigating information dilution, and reducing hallucinations in engineering-oriented applications, this paper proposes SPR-RAG, a retrieval-augmented framework designed for knowledge-intensive vertical domains such as electric power policy analysis. With practicality and interpretability as core design goals, SPR-RAG introduces a Semantic Parsing Retriever (SPR), which integrates community detection–based entity disambiguation and transforms natural language queries into logical forms for structured querying over a domain knowledge graph, thereby retrieving verifiable triple-based evidence. To further resolve retrieval bias arising from diverse policy writing styles and inconsistencies between user queries and policy text expressions, a question-repository–based indirect retrieval mechanism is developed. By generating and matching latent questions, this module enables more robust retrieval of non-structured contextual evidence. The system then fuses structured and unstructured evidence into a unified dual-source context, providing the generator with an interpretable and reliable grounding signal. Experiments conducted on real electric power policy corpora demonstrate that SPR-RAG achieves 90.01% faithfulness—representing a 5.26% improvement over traditional RAG—and 76.77% context relevance, with a 5.96% gain. These results show that SPR-RAG effectively mitigates hallucinations caused by ambiguous entity names, textual redundancy, and irrelevant retrieved content, thereby improving the verifiability and factual grounding of generated answers. Overall, SPR-RAG demonstrates strong deployability and cross-domain transfer potential through its “Text → Knowledge Graph → RAG” engineering paradigm. The framework provides a practical and generalizable technical blueprint for building high-trust, industry-grade question–answering systems, offering substantial engineering value and real-world applicability. Full article

This paper forwards the claim that our early human ancestors had protosacred experiences long before they had languages, architecture, or religions. A mountain may create feelings of awe while a grove in the forest may create feelings of serenity. In some circumstances (and very much dependent on the mental set of the individual), such protosacred experiences may create a sense of ultimacy that may be interpreted by the faithful as a religious experience in terms of their own beliefs. We chart an evolutionary account of the path of human ancestors from experiences of the protosacred to the diversity of religions, with a focus on the emergence of culturally varied architected sacred spaces designed to offer a religious group a sense of shared community and the sacred in the experience of their religion. We argue that the cultural evolution of languages was necessary for this transition. It made our species both Homo quaerens (the humans who ask questions) and Homo narrans (the humans who tell stories), able to ask existential questions and to offer answers that a group could accept. The answers took the form of narratives and scripts for ritual behaviors that could harmonize the community with the world around and beyond it. We suggest that both affordances and atmospheres relate to the aesthetics of space, stressing the atmospheric flow as the performance of various rituals proceeds. This paper offers examples from diverse religions or cosmologies and closes with suggestions for a range of empirical and experimental investigations to address the hypotheses raised herein. Full article

As artificial intelligence (AI) reshapes production, its integration into manufacturing offers gains in precision, efficiency, and sustainability. Globally, AI supports additive, subtractive, and forming processes through optimization, monitoring, defect detection, and design innovation. In Latin America, however, adoption is limited and uneven, with most evidence from surveys, policy reports, and pilot projects rather than large-scale implementations. This review addresses that gap by examining the global landscape of AI in manufacturing and the specific conditions influencing its adoption in the region. The study is guided by the question: What structural conditions are required to enable successful and sustainable AI integration in Latin American manufacturing? To answer, it applies the Triadic Integration Framework, which identifies three pillars: digital infrastructure, policy and governance, and socio-industrial capacity. The analysis highlights barriers, including fragmented regulation, skills shortages, cybersecurity risks, and cost–benefit uncertainties, while also pointing to opportunities in various industrial sectors. To translate insights into practice, a phased roadmap is proposed, outlining short-term, medium-term, and long-term actions, along with the responsible stakeholders and the necessary resources. As an integrative review, the study synthesizes existing knowledge to build a framework, defining directions for future research, emphasizing that successful adoption requires technical progress, inclusive governance, and regional coordination. Full article

With the advancement of artificial intelligence, Agricultural Expert Systems (AESs) show great potential in enhancing agricultural management efficiency and resource utilization. Accurate extraction of semantic features from agricultural short texts is fundamental to enabling key functions such as intelligent question answering, semantic retrieval, and decision support. However, existing single-structure deep neural networks struggle to capture the hierarchical linguistic patterns and contextual dependencies inherent in domain-specific texts. To address this limitation, we propose a hybrid deep learning model—Bidirectional Encoder Recurrent Neural Network (BERNN)—which combines a domain-specific pre-trained Transformer encoder (AgQsBERT) with a Bidirectional Long Short-Term Memory (BiLSTM) network. AgQsBERT generates contextualized word embeddings by leveraging domain-specific pretraining, effectively capturing the semantics of agricultural terminology. These embeddings are then passed to the BiLSTM, which models sequential dependencies in both directions, enhancing the model’s understanding of contextual flow and word disambiguation. Importantly, the bidirectional nature of the BiLSTM introduces a form of architectural symmetry, allowing the model to process input in both forward and backward directions. This symmetric design enables balanced context modeling, which improves the understanding of fragmented and ambiguous phrases frequently encountered in agricultural texts. The synergy between semantic abstraction from AgQsBERT and symmetric contextual modeling from BiLSTM significantly enhances the expressiveness and generalizability of the model. Evaluated on a self-constructed agricultural question dataset with 110,647 annotated samples, BERNN achieved a classification accuracy of 97.19%, surpassing the baseline by 3.2%. Cross-domain validation on the Tsinghua News dataset further demonstrates its robust generalization capability. This architecture provides a powerful foundation for intelligent agricultural question-answering systems, semantic retrieval, and decision support within smart agriculture applications. Full article

This comment focuses on the contribution of experimental brain ischemia to the overwhelming incidence of Alzheimer’s disease in women as presented by Lohkamp et al. in Life 2025, 15, 333. The authors showed that in Alzheimer’s disease and ischemic stroke there are sex-dependent adaptations in the form of cross-links and vice versa. It was emphasized that the high longevity of women in itself does not explain the mechanisms underlying the biological differences between the sexes causing a female predominance in the development of Alzheimer’s disease. Differences were demonstrated between males and females: female APP/PS1 mice had greater amyloid deposition, hyperactivity, lower body weight, and reduced cerebral blood flow, as well as less neuroinflammation, which the authors suggest may have potential neuroprotection. It should be noted that some of the information presented in the article by Lohkamp et al. raises more questions than answers. Therefore, future studies should consider, for example, studies using single-cell technologies that can provide insight into the timing and sequence of cellular dysfunctions across sexes and analyze the continuity of changes over time, starting from short-term observations of a few days and ending with long-term observations of a year or more, to assess the continuity and differentiation of changes. Full article

Fusion-in-Decoder (FiD), a prominent retrieval-augmented generation model, has demonstrated outstanding performance in open-domain question answering by effectively leveraging multiple passages. However, processing multiple passages significantly increases computational costs at both encoder and decoder components. In particular, in Long-Form Question Answering (LFQA) scenarios, the decoder’s cross-attention computation scales proportionally with the length of the generated answer, severely impacting the overall inference speed. In this paper, we propose a novel dynamic token pruning mechanism to alleviate the computational bottleneck of the FiD decoder. Our method selectively identifies and removes tokens predicted to have low contributions to answer generation by jointly considering their contextual information and attention scores within the FiD encoder. The resulting pruned representations are then passed to the decoder, significantly reducing the cross-attention computations and thereby accelerating the overall inference process. Experimental evaluations on two LFQA benchmarks, ASQA and CLAPNQ, demonstrate that the proposed method achieves up to a 1.74-fold speed-up while maintaining minimal degradation in answer quality, effectively enhancing computational efficiency compared to the original FiD model. Full article

The objective of ensuring equal access to and enjoyment of tourism for the broadest spectrum of individuals, regardless of age or ability, is a fundamental right for all, as explicitly outlined in the United Nations Convention on the Rights of Persons with Disabilities. Nevertheless, notwithstanding the initiatives aimed at actualizing the aims and objectives of the Convention, the discrepancy between the supply and demand for accessibility remains considerably high in Italy. In fact, numerous accessibility issues persist in information, services, transportation, tourist destinations, accommodations, and various types of facilities and attractions. The inadequacy of long-term planning and the lack of a comprehensive perspective on accessibility further exacerbate the situation in Italy. In light of these considerations, this paper aims to examine, via a survey conducted on a sample of potential users, the challenges and opportunities for the development of inclusive forms of accessible tourism and to recommend a reference framework for best practices that encompasses, in addition to barriers, the elements that enhance accessibility and usability of cultural activities for individuals with disabilities, serving as a reference point to assist in the planning and governance of sustainable tourism policies. This paper aims to answer three fundamental research questions to improve the conditions of the Italian tourism system: What is the perception of people with disabilities regarding their ability to travel in Italy? What needs to be improved to achieve a good tourism experience for people with disabilities in Italy? Which parameters should be considered for proper planning of accessible and inclusive tourism in Italy? Full article

A recent analysis compared the proteome of (i) blood clots seen in two diseases—sepsis and long COVID—when blood was known to have clotted into an amyloid microclot form (as judged by staining with the fluorogenic amyloid stain thioflavin T) with (ii) that of those non-amyloid clots considered to have formed normally. Such fibrinaloid microclots are also relatively resistant to fibrinolysis. The proteins that the amyloid microclots contained differed markedly both from the soluble proteome of typical plasma and that of normal clots, and also between the diseases studied (an acute syndrome in the form of sepsis in an ITU and a chronic disease represented by Long COVID). Many proteins in the amyloid microclots were low in concentration in plasma and were effectively accumulated into the fibres, whereas many other abundant plasma proteins were excluded. The proteins found in the microclots associated with the diseases also tended to be themselves amyloidogenic. We here ask effectively the inverse question. This is: can the clot proteome tell us whether the clots associated with a particular disease contained proteins that are observed uniquely (or are highly over-represented) in known amyloid clots relative to normal clots, and thus were in fact amyloid in nature? The answer is in the affirmative in a variety of major coagulopathies, viz., venous thromboembolism, pulmonary embolism, deep vein thrombosis, various cardiac issues, and ischaemic stroke. Galectin-3-binding protein and thrombospondin-1 seem to be especially widely associated with amyloid-type clots, and the latter has indeed been shown to be incorporated into growing fibrin fibres. These may consequently provide useful biomarkers with a mechanistic basis. Full article

Following Einstein’s prediction that “Physics constitutes a logical system of thought” and “Nature is the realization of the simplest conceivable mathematical ideas”, this topical review outlines a formal extension of local realism limited by the speed of light to global realism with bipolar strings (GRBS) that unifies the principle of locality with quantum nonlocality. The related literature is critically reviewed to justify GRBS which is shown as a necessary and inevitable consequence of the Bell test and an equilibrium-based axiomatization of physics and quantum information science for brain–universe similarity and human-level intelligence. With definable causality in regularity and mind–light–matter unity for quantum superposition/entanglement, bipolar universal modus ponens (BUMP) in GRBS makes quantum emergence and submergence of spacetime logically ubiquitous in both the physical and mental worlds—an unexpected but long-sought simplification of quantum gravity with complete background independence. It is shown that GRBS forms a basis for quantum intelligence (QI)—a spacetime transcendent, quantum–digital compatible, analytical quantum computing paradigm where bipolar strings lead to bipolar entropy as a nonlinear bipolar dynamic and set–theoretic unification of order and disorder as well as linearity and nonlinearity for energy/information conservation, regeneration, and degeneration toward quantum cognition and quantum biology (QCQB) as well as information-conservational blackhole keypad compression and big bang data recovery. Subsequently, GRBS is justified as a real-world quantum gravity (RWQG) theory—a bipolar relativistic causal–logical reconceptualization and unification of string theory, loop quantum gravity, and M-theory—the three roads to quantum gravity. Based on GRBS, the following is posited: (1) life is a living bipolar superstring regulated by bipolar entropy; (2) thinking with consciousness and memory growth as a prerequisite for human-level intelligence is fundamentally mind–light–matter unitary QI logically equivalent to quantum emergence (entanglement) and submergence (collapse) of spacetime. These two posits lead to a positive answer to the question “If AI machine cannot think, can QI machine think?”. Causal–logical brain modeling (CLBM) for entangled machine thinking and imagination (EMTI) is proposed and graphically illustrated. The testability and falsifiability of GRBS are discussed. Full article

Background: Long COVID refers to the persistence or development of signs and symptoms well after the acute phase of COVID-19. Objective of the study: To investigate the long-term outcomes of the SARS-CoV-2 infection in terms of psychological, social, and relational consequences within the Italian population. Materials and methods: We conducted an observational, cross-sectional, and multicenter study using an online questionnaire distributed to a sample of the Italian population. By utilizing the Short Form 12 Health Survey (SF-12) and the Hikikomori scale, we assessed perceived quality of life and social isolation, respectively. The questionnaire also included an open-answer question: “What will you remember about the pandemic period?”. We used generative artificial intelligence to analyze and summarize the corresponding answers. Results: A total of 1097 people participated in this study. A total of 79.3% (n = 870) of participants declared that they had been hospitalized and 62.8% (n = 689) received home care. Physical symptoms included headaches (43%, n = 472) and asthma (30.4%, n = 334). Additionally, 29.2% (n = 320) developed an addiction during the pandemic and, among these, 224 claimed internet addiction while 73 declared an emotional addiction. Furthermore, 51.8% (n = 568) experienced limitations in carrying out daily life activities. According to the Hikikomori scale, participants with positive SARS-CoV-2 infection exhibited higher levels of isolation compared to the others (p < 0.001). Participants without COVID-19 showed higher levels of emotional support (p < 0.001). Our semiautomatic analysis of the open-ended responses, obtained by a procedure based on a free large language model, allowed us to deduce and summarize the main feelings expressed by the interviewees regarding the pandemic. Conclusions: The data collected emphasize the urgent need to investigate the consequences of long COVID in order to implement interventions to support psychological well-being. Full article

Urban agriculture refers to any type of activity located within or around a city designed to provide ecosystem services. Given the rapid population growth and urbanization, urban agriculture is seen as a potential alternative route to a more sustainable urban food system. This review answers the main question: What are the benefits of non-commercial of Urban Agriculture (NCUA) forms and its contribution towards food production? using a systematic literature review approach. The methodology involved capturing 1355 recent articles from qualified search engines, using key terms according to the defined question, then screened for relevance and the defined scope of this review, resulting in a final selection of 40 articles for analysis. The results show that implementing NCUA practices has multifaced social, economic, and environmental benefits, such as improving people’s health, reducing expenditure on food and creating sustainable cities, highlighting the need to recognize the multifaceted role of NCUA in promoting a more sustainable lifestyle and strengthening local communities and engagement. Moreover, awareness of urban agriculture differs between developed and developing countries, as does the recognition and valorization of its benefits. Further research is needed to examine the enabling factors and barriers to NCUA adoption in different urban context, the resource implications, and the long-term sustainability of these practices. Full article

School inclusion is based on the need to adopt and implement a holistic view of education, training, and human development embodied in the idea of everyone, for everyone. In the context of Physical Education (PE), there are still several constraints to the realization of this universal desideratum. Among these, teacher training and qualification for the inclusion of students with Specific Health Needs (SHNs) stands out. That is, students with physical and mental health problems whose impact is significantly manifested in the learning process. Thus, the objective of this study was to identify the representations of PE teachers about their training to develop inclusive processes with students with SES. Participants in this study were 151 PE teachers from different regions and districts of Portugal (Algarve, Aveiro, Castelo Branco, Lisbon, Porto, and Viseu) who had 23.6 ± 8.1 years of teaching service. Teachers answered an online questionnaire, on the Google Forms platform, with open and closed questions about their education and training to develop inclusive processes in PE. The results indicate two significant dimensions: (1) initial training for teaching inclusive PE and (2) continuous training for inclusion. Regarding initial training, a large majority of the teachers under study, at the end of their initial training, did not have the essential skills to teach PE to students with SES. It was also identified that a large majority reported not having had any contact with students with SES throughout their training process for teaching. It was also recognized that this training was not adjusted to the development of intervention skills with students with SHN. Regarding continuous training, it was identified that attendance at this training increased their skills to teach PE to students with SHN. Workshops/actions/training courses are the main training models adopted. However, it is recognized that the training provided does not respond concretely to their training needs to intervene with students with SHN, since teachers essentially seek to improve intervention in the context of inclusive physical education. We conclude that teacher training for inclusion is not yet fully adjusted to the reality of the inclusive school paradigm. In this sense, in practical terms, the following are suggested: (1) the need for reinforcement in study plans with specific and long-term curricular units; (2) the introduction of real practice components in context; and (3) supervised pedagogical practice in diverse contexts. Full article

Current State-of-the-Art (SotA) chatbots are able to produce high-quality sentences, handling different conversation topics and larger interaction times. Unfortunately, the generated responses depend greatly on the data on which they have been trained, the specific dialogue history and current turn used for guiding the response, the internal decoding mechanisms, and ranking strategies, among others. Therefore, it may happen that for semantically similar questions asked by users, the chatbot may provide a different answer, which can be considered as a form of hallucination or producing confusion in long-term interactions. In this research paper, we propose a novel methodology consisting of two main phases: (a) hierarchical automatic detection of topics and subtopics in dialogue interactions using a zero-shot learning approach, and (b) detecting inconsistent answers using k-means and the Silhouette coefficient. To evaluate the efficacy of topic and subtopic detection, we use a subset of the DailyDialog dataset and real dialogue interactions gathered during the Alexa Socialbot Grand Challenge 5 (SGC5). The proposed approach enables the detection of up to 18 different topics and 102 subtopics. For the purpose of detecting inconsistencies, we manually generate multiple paraphrased questions and employ several pre-trained SotA chatbot models to generate responses. Our experimental results demonstrate a weighted F-1 value of 0.34 for topic detection, a weighted F-1 value of 0.78 for subtopic detection in DailyDialog, then 81% and 62% accuracy for topic and subtopic classification in SGC5, respectively. Finally, to predict the number of different responses, we obtained a mean squared error (MSE) of 3.4 when testing smaller generative models and 4.9 in recent large language models. Full article

Arguing and communicating are basic skills in the mathematics curriculum. Making arguments in written form facilitates rigorous reasoning. It allows peers to review arguments, and to receive feedback about them. Even though it requires additional cognitive effort in the calculation process, it enhances long-term retention and facilitates deeper understanding. However, developing these competencies in elementary school classrooms is a great challenge. It requires at least two conditions: all students write and all receive immediate feedback. One solution is to use online platforms. However, this is very demanding for the teacher. The teacher must review 30 answers in real time. To facilitate the revision, it is necessary to automatize the detection of incoherent responses. Thus, the teacher can immediately seek to correct them. In this work, we analyzed 14,457 responses to open-ended questions written by 974 fourth graders on the ConectaIdeas online platform. A total of 13% of the answers were incoherent. Using natural language processing and machine learning algorithms, we built an automatic classifier. Then, we tested the classifier on an independent set of written responses to different open-ended questions. We found that the classifier achieved an F1-score = 79.15% for incoherent detection, which is better than baselines using different heuristics. Full article