Scholarly knowledge graphs integrate bibliographic records from heterogeneous sources and therefore require controlled, auditable deduplication. This paper presents OntoDup, an ontology-driven approach that models entity matching as a governed decision process: Matching outcomes are recorded as reified assertions enriched with governance state, evidence, provenance and operational metadata, while a separate operational view is exposed through policy-driven materialization of consumable identity links. We evaluate OntoDup on the DBLP-ACM and DBLP-Scholar benchmarks under two regimes: (i) a pre-blocked setting using the benchmark candidate lists to compare matching methods under a fixed candidate set, and (ii) an end-to-end setting that generates candidates from the graph with DeepBlocker and applies governed triage and materialization. We report operational precision/recall/F1 computed directly on the graph via SPARQL aggregations, characterize governance workload through state distributions, and quantify inference cost for LLM-based matchers via token and latency metadata attached to assertions. For end-to-end evaluation, we anchor operational links against a full positive reference encoded as idealized validations derived from the benchmark labels, enabling analysis of missed positives in terms of governance status and materialization policy. The experiments show that OntoDup enables evaluation at the level of consumable identity links, review workload, and inference cost, revealing operational trade-offs that are not visible from pairwise matching metrics alone.

Scientific collaboration is increasingly needed to address complex research challenges, yet identifying promising partners in the absence of prior co-authorship remains difficult. We present a decision-support pipeline for discovering researchers who have not previously worked together and whose collaboration is unlikely to emerge without deliberate intervention or institutional incentives. The approach leverages document-level semantic representations to estimate proximity between publications, aggregates these similarities at the author level, and surfaces collaboration opportunities that are not evident from the co-authorship graph. To support interpretation by decision makers, a separate LLM module proposes potential joint research directions, which are subsequently annotated with multi-label fields of study. We evaluate the pipeline through an institutional case study, analyzing 7531 publications from 2009 to 2024 using retrospective, temporally shifted windows. While only a small fraction of suggested pairs materialized spontaneously in subsequent periods, the collaborations that do emerge exhibit strong semantic alignment with the computed recommendations (high cosine similarity) and substantial thematic overlap. These results indicate that semantic proximity can act as an early indicator of latent complementarity between researchers without prior ties, supporting intentional institutional mediation and complementing topology-driven approaches that predict links under passive evolution.

This paper examines whether research-space analysis can support portfolio design informed by forecasts at the country–subfield level. Using a bibliographic catalog of scientific papers, authors and institutions, we construct a research space from country activity across subfields and test whether measures derived from that space help predict later changes in specialization. The study contributes an empirical framework that links three elements: historical validation of research-space measures, comparison of alternative specifications of relative relatedness, and portfolio selection under an explicit complexity target. We estimate reduced-form stepping-stone models for entry and exit and evaluate them using in-sample evidence, leave-one-year-out checks, and robustness comparisons across relatedness constructions. The results show that research-space measures help predict later specialization changes and that relative relatedness standardized over the opportunity set provides the most consistent specification, especially for entry into previously non-specialized subfields. Using this specification, we formulate a prospective portfolio problem in which candidate subfields are selected to increase portfolio complexity while minimizing predicted entry effort. In a 2024 comparison covering 164 countries, optimized portfolios differ markedly from relatedness rankings, with 92.7% of countries shifting toward lower auxiliary volume and higher predicted future sophistication.

Topic analysis within broad and evolving fields poses great challenges when attempting to be addressed by traditional methods. In response, topic modeling seeks to automate the identification and analysis of underlying themes within several collections of documents in order to synthesize and facilitate the interpretation of their content. The present study applied multiple iterations of the Latent Dirichlet Allocation (LDA) model and the Best-K mechanism in the identification of topics within a volume of 250 news items labeled with “Artificial Intelligence” within a mainstream web portal. Additionally, a large language model (LLM) was implemented to improve interpretation and description for the labeling of found topics. As a result, 9 key topics were identified ranging from the main trends and challenges in AI.

In just one decade, Mentalization-Based Therapy (MBT) has progressed from the gold-standard intervention for borderline personality disorder (BPD) to an increasingly consolidated transdiagnostic framework. To chart this trajectory, we analysed 405 publications indexed in Scopus and Web of Science (2015-2024). Bibliometric processing with Bibliometrix and VOSviewer revealed an uninterrupted growth curve-compound annual growth rate $\approx 8 \%$-driven by a collaborative network dominated by the United Kingdom, the United States and Nordic hubs. Eight core investigators, spearheaded by Fonagy and Bateman, account for $33 \%$ of the literature, while openaccess journals publish one-third of the output, accelerating clinical translation. Thematically, MBT research still gravitates around BPD; however, keyword co-occurrence networks signal a decisive migration toward eating disorders, forensic contexts and early psychosis. This bibliometric mapping shows that MBT is expanding in synchrony with health-service demands for scalable, cost-efficient psychological interventions and underscores the need for multicentre trials to validate its effectiveness across diverse cultural and service settings.

This study traces how Ecuadorian news outlets framed the 2025 general elections on social platforms. We analysed 8174 posts from 48 media accounts on Facebook, Instagram, TikTok and X (1 Jan-29 Jun 2025). Spanish texts were cleaned, embedded with a RoBERTa model, reduced via UMAP and clustered with BERTopic-HDBSCAN. Thirty-one topics emerged (mean coherence $\mathbf{C}-\mathbf{v}=\mathbf{0. 5 0}$; diversity $=\mathbf{0. 8 4}$); five hold 47.5% of the corpus and centre on Daniel Noboa and territorial voting patterns. Posting spikes on 9 Feb, 23 Mar and 13 Apr coincide with the first round, the presidential debate and the runoff, confirming an event-driven agenda. A document map shows dense central clusters for high-volume themes and scattered points for marginal narratives. Content production remains concentrated: three outlets generate nearly one-third of posts and Facebook carries over half the traffic, suggesting that social-platform affordances have not displaced traditional logics of centralisation and candidate-centric framing in Ecuadorian electoral journalism.

This paper explores the development of an Intelligent Educational Agent (IEA) with a focus on enhancing the learning experience for university students. In an era where online education is on the rise, there is a growing demand for personalized learning tools. IEAs, powered by artificial intelligence, offer a solution by providing tailored support, explanations, answers to queries, and content adaptation. This study leverages advanced AI technologies, including the LangChain framework and the GPT-3.5 Turbo model from OpenAI, to create an adaptive educational assistant. LangChain facilitates Natural Language Processing and information analysis, while GPT-3.5 Turbo ensures context-aware responses through prompt-tuning. The research methodology involves defining functional requirements, implementing the LangChain framework for NLP, integrating the OpenAI API, and establishing an architecture with three main actors: students, teachers, and tutors/assistants. Results indicate the IEA’s ability to generate precise multiple-choice tests and comprehensive academic plans. The system exhibits contextual understanding and resource generation capabilities. In conclusion, despite challenges like data quality and infrastructure requirements, developing an IEA for content adaptation based on large language models shows great promise. It has the potential to revolutionize education by providing personalized learning experiences and generating educational resources. Collaboration among education experts, developers, and researchers is crucial to fully harness this transformative potential.

In the contemporary information age, knowledge extraction from vast textual datasets become essential. Named Entity Recognition (NER) models emerge as fundamental tools for this task, focusing on identifying key elements, i.e., entities. This study is based on a 'spaCy's-es_core_news_lg' NER model focusing on the geo-positioning of entities in publications related to security analysis in Ecuador. During increasing violence in the country, this work aims to improve situational awareness through NER models, enabling agile and effective responses from authorities. Geo-analysis of publications from the social network X (Twitter) is used, with a model that facilitates understanding through graphs, visualizing the distribution of security cases and violence in various sectors of Ecuador. The methodology adopted uses a modular pipeline, prioritizing the cleaning of text to enhance the precision of the results. The creation of density maps and spatial trend analysis supports the application of NER in the geolocation of entities. These results are anticipated to improve performance in critical areas such as trend analysis, information extraction, and thematic labeling, strengthening the ability to make informed decisions in crucial situations.

According to latest estimates of the World Health Organization (WHO), now a days there are 285 million people visually impaired world wide: 39 million are blind, 246 have low vision, and 19 million are children. This situation becomes more complex in developing countries because the local governments and families do not have enough resources and assistive technologies to improve the literacy programs of blind people. For these reasons, in this paper, we present an integral low-cost tool to support the learning process of Braille code at schools or homes. The novelty of our proposal is that it relies on a multi-layer architecture that provides the following functionalities: support to carry out reading and writing exercises using a single device, an interactive interface for designing and monitoring classroom activities through an approach based on data mining and rule-based reasoning, and a mobile application to interact with the system easily. Our proposal has been evaluated in a pilot experiment by 20 persons: 7 (blind) teachers that work with visual impaired/blind children, and 13 persons with normal vision. The results achieved during the evaluation of our proposal are encouraging: 90% of participants agree with the usefulness of the system as well as the high potential of including the tool in classrooms.

This chapter presents a new proposal for supporting the management of research processes in universities and higher education centers.To this aim, the authors have developed a comprehensive ecosystem that implements a knowledge model that addresses three innovative aspects of research: (i) acceleration of knowledge production, (ii) research valorization and (iii) discovery of improbable peers.The ecosystem relies on ontologies and intelligent modules and is able to automatically retrieve information of major scientific databases such as SCOPUS and Science Direct to infer new information.Currently, the system is able to provide guidelines to create improbable research peers as well as automatically generate resilience graphics and reports from more than 17,000 tuples of the ontological database.In this work, the authors describe in detail an important aspect of support systems for research management in higher education: the development and valorization of competences of students collaborating in research process and startUPS of universities.Furthermore, a knowledge model of entrepreneurship (startUPS) as well as an analyzer of general and specific competences based on data mining processes is presented.

According to latest estimates of World Health Organization (WHO), nowadays 93 million of children live with severe or moderate disabilities. In the same way, these studies point that around 15% of world population present some form of disability. In addition to this complex situation, it is fundamental to consider that an important number of elderly are developing different kinds of disabilities. From this group, a significant percentage of persons present different types of communication disorders that can limit their lives in different ways. Given that one of most vulnerable groups is constitued by children, in this paper we present a low-cost robotic assistant able to support their rehabilitation process through Speech-LanguageTherapy (SLT). This robotic assistant is able to interact with a mobile application that is aimed to conduct reinforcement activities at home with the parents of the children. Our proposal has been tested with 29 children with cerebral palsy and communication disorders. The results are encouraging, given that it was possible to success fully integrate the robot to the therapy sessions.

Segun las ultimas estimaciones de la Organizacion Mundial de la Salud (OMS), en la actualidad 93 millones de ninos viven con discapacidades severas o moderadas. Del mismo modo, estos estudios apuntan que alrededor del 15% de la poblacion mundial presenta algun tipo de discapacidad. Ademas de esta compleja situacion, es fundamental considerar que un numero importante de ancianos estan desarrollando diferentes tipos de discapacidades. De este grupo, un porcentaje considerable de personas presentan distintos tipos de trastornos de la comunicacion que pueden limitar sus vidas de diferentes maneras. Dado que uno de los grupos mas vulnerables esta constituido por ninos, en este articulo se presenta un asistente robotico de bajo costo que es capaz de apoyar su proceso de rehabilitacion a traves de la Terapia del Habla y el Lenguaje (THL). Este asistente robotico es capaz de interactuar con una aplicacion movil que tiene como objetivo llevar a cabo actividades de refuerzo en el hogar con los padres de los ninos. Nuestra propuesta ha sido puesta a prueba con 29 ninos con paralisis cerebral y trastornos de la comunicacion. Los resultados obtenidos son alentadores, dado que fue posible integrar al robot de forma exitosa a las sesiones de terapia.

This research develops an informatics application, consisting in four virtual games that present three situations in which trauma produced by accidents in children occur. The games contain preventive information, supported by graphics which describe secure behavior, possible consequences of accidents and some elements that the child needs to avoid, immerse in children's common sceneries as home and school. A familiar character (dog) was implemented to create a reference of security in children, so they could be able to associate all the information presented, in one figure. The tool was tested with two groups of regular school students from third to fifth grade. 31 children assisted to a traditional workshop, including physical ludic activities to transmit preventive information about accidents, and 30, learnt through the informatics application. Finally, students were evaluated through a brief questionnaire. In a preliminary analysis, results showed a higher impact on the group that used the informatics application, overpassing the traditional workshop with 6% points. To verify this, the results of the different topics of the workshops: pedestrian accidents, burns and intoxication, were averaged, evidencing a precision of 73.14% for the traditional workshop, and 79.72% for the one using the application. This result will be complemented in the future, with a second analysis of variables, to determine the relationship between the elements of the virtual games and the effectiveness of the achieved learning.

According to the latest estimates of the World Bank, up to 2013 there were 2'184.419.897 scientific and technical articles published in the world. Year in and year out, a great number of these papers is produced by several education institutions, organizations and centers that carry out several research topics and processes. However, it is imperative for universities to determine guidelines and policies that efficiently enable them to manage the information that has been produced, as well as determine the impact all of these processes have on several areas, such as scientific, social and academic. Therefore, this article presents an ecosystem based on ontologies and smart tools that dynamically analyzes scientific production in a university. Our idea is capable of connecting to main scientific data bases such as SCOPUS or Science Direct and applies standards like the International Standard Classification of Education from UNESCO and the All Science Journal from SCOPUS. In order to validate our proposal, we carried out an experimental process that allowed us to supply the ontology with 1.912 individuals of papers, authors, journals and organizations. Similarly, from the 331 articles that have been processed, we have extracted information related to lines of research, keywords, scientific events and many other fields.

Nowadays the World Health Organization (WHO) claims that one billion persons live in the world with some form of disability. From this group, an important number of children with disabilities and communication disorders do not have access to adequate services for special education or health care. In this point, it is important to mention that language and communication are mainstays in the development of several intellectual, social, and cognitive human skills. Therefore, in this paper we present a low-cost robotic assistant that is able to provide support for several activities that must be carried during the speech-language therapy sessions of children with different kinds of disabilities. This assistant is able to register patient's information (personal and clinical data), the results of therapy sessions, and provide remote support to conduct reinforcement activities at patient's home (using mobile devices). With the aim of determining the real feasibility of our system, we have carried out a pilot experiment in 73 therapy sessions with 29 children with disabilities (15 received traditional therapy and 14 received therapy with the assistance of the robot). The results show a rapid adaptation of children to the new tool and good results in phonological, morphosyntactical, and semantic areas.