As surveillance cameras become commonly used, the demand for automatic violence detection technology increases. A significant number of works have been developed in the area of automatic violence detection, utilizing a diverse range of approaches in different scenarios. The scenarios explored by researchers usually revolve around violence detection in sports, movies, and outdoor spaces, but works that involve violence detection against women are still way too few. Violence against women is a pervasive global public health issue, with alarming statistics indicating an increasing trend over time. Domestic violence is a complex multifaceted problem, in which no single strategy can address the problem by itself. This work introduces a technological tool, SUSAN (Surveillance System Architecture based on deep neural Networks), an architecture designed for detecting violence against women in surveillance videos. SUSAN employs deep learning models for human detection, violence detection, and gender classification. In this work, one model for human detection was fixed, and four models of Convolutional Neural Networks were combined in pairs for the tasks of violence detection and gender classification and evaluated with four metrics: accuracy, precision, recall, and F1-score. This work utilized the AIRTLab and PA-100K datasets for training and the first for testing the architecture as it is composed of synthetic violent clips involving women. The proposal presented promising results, with the best combination of models reaching 73% of accuracy, 80% of F1-score, 78% of precision and 82% of recall, which demonstrates the viability of an algorithm for automatic violence detection against women, and contributed to a research area still in development.

This demo paper presents SIMBA, a web-based tool designed to support instructors in creating learning activities with LLM-powered chatbots. It functions as a Socratic tutor, fostering students' metacognitive skills through structured questioning. By defining learning objectives and a sequence of guiding questions, instructors generate customized chatbots that act as learning companions. These can be shared with students, while instructors monitor interactions via an interactive analytics dashboard. SIMBA was evaluated during a workshop with eight instructors and the results suggest that SIMBA holds promise as a scaffolding tool for active learning, leveraging LLMs to enhance metacognitive engagement while providing instructors with actionable insights.

Large Language Models (LLMs) have rapidly gained popularity among university students for tasks such as essay writing, virtual tutoring, and coding exercises. Although the number of studies applying LLM-based solutions in educational contexts has grown significantly, researchers emphasize the need for empirical studies to evaluate the effectiveness of these tools to support teaching and learning. To address this gap, this paper presents a study analyzing the effects of an LLM-based system designed to support students' self-reflection. The study compares the grades and course outcomes of two groups of students enrolled in the same Thermodynamics course during two academic years (2023 and 2024, 234 students in total), but only the 2024 cohort had access to an LLM-based chatbot specifically designed to support self-reflection. The data showed no significant correlation between students' AI literacy profiles, their prior experience with generative AI, and the adoption of the tool. However, an analysis of engagement with the tool in the 2024 cohort revealed that students who interacted more extensively with the chatbot particularly medium and high achievers (based on prior academic performance) demonstrated a significant improvement in their final grades.

Within the spectrum of autism, there are children who are unable to communicate orally, which places severe barriers to their education and integration to society. Augmentative and Alternative Communication (AAC) technologies assist these children in expressing themselves, usually through a card-based interface. AAC apps became very popular in schools, helping educators to include children with disabilities in learning activities. In this context, new possibilities that go beyond communication emerge from the teachers and students’ creativity, in particular the creation of playful educational content using these apps. However, as AAC apps were not designed for this goal, there are many limitations when users attempt to make these adaptations. This situation motivated the experience presented in this paper, where we worked with teachers and children with autism who already used the Livox AAC app, to design a new module with the aim of providing the necessary tools for creating games in the context of alternative communication, while still maintaining accessibility. The experience reinforced the importance to think about alternative methods of co-design and be ready to adapt when co-designing with children with autism, so that they can effectively contribute to the design process, having their opinions, wishes and capabilities truly taken into account.

The rapid evolution and widespread accessibility of non-invasive medical imaging technologies, exemplified by Magnetic Resonance Imaging (MRI) and Computerized Tomography (CT), are fundamentally reshaping medical decisionmaking paradigms. These sophisticated imaging modalities, capable of extracting high-definition medical images, have emerged as integral components of modern healthcare, facilitating precise diagnostics and treatment planning. The escalating adoption of such technologies, however, has accentuated the need for a nuanced understanding and optimization of the performance and productivity of both medical equipment and the teams operating them, mainly due to the high costs and risks caused by their misuse. This work proposes using univariate analytical models to estimate the number of exams performed per day with machine learning algorithms. For such, different energy-related sensors monitoring 25 magnetic resonance equipment from three different brands were considered. The results of the research reveal a compelling validation of the proposed approach. A notably high Pearson correlation coefficient is observed between the predictions generated by the evaluated models and the real measurements obtained through the Radiology Information System (RIS). This robust correlation emphasizes the accuracy and reliability of the estimation models, validating their potential applicability in real-world healthcare scenarios. Furthermore, the study unveils an intriguing trend that distinguishes the performance of electric current sensors. Thirteen out of the 25 evaluated MRI machines demonstrate superior results when equipped with electric current sensors compared to other sensor types. This nuanced insight not only substantiates the critical role of energy-related sensors in predicting equipment performance but also underscores the importance of tailoring monitoring strategies to the unique characteristics of each machine.

Accurate punctuation in written text enables unambiguous communication, minimizing the risk of misunderstandings. Conversely, faulty punctuation can confuse the intended meaning, posing challenges for the author. The existing literature offers a collection of systems and algorithms to assist users in writing tasks. However, those focusing on English tend to exhibit higher accuracy. Furthermore, most models for punctuation restoration yield results without offering insight into their decision-making processes. Therefore, this study evaluated state-of-the-art punctuation restoration models specifically for Brazilian Portuguese and incorporated the principles of explainable artificial intelligence to clarify their predictions transparently. The findings indicate that the models assessed achieved an accuracy comparable to those of their English-language counterparts.

Narrative essays enable students to express their thoughts and feelings, gain diverse perspectives, understand themselves and their world more deeply, and enhance their literary skills and cultural awareness. As such, it became a relevant textual production in educational settings. However, assessing these texts is challenging and time-intensive. Therefore, this study focuses on developing an automatic detection system for narrative structures in essays, using a range of natural language processing algorithms, including machine learning, deep learning, and large language models. The study found that BERT was more effective than other models for this task, highlighting GPT's potential in extracting narrative structures. The findings of this study have implications for educational practices, particularly in the assessment and improvement of narrative writing skills among middle school students.

Texts produced by the Brazilian judiciary have a complex and technical vocabulary, with elaborate use of the Portuguese language and many legal terms difficult to be understood, generating a barrier in communication between the judiciary and the population. In this sense, the Automatic Text Simplification (ATS), activity of the Natural Language Processing (NLP) area, can be applied to improve the readability of these types of text using specialized algorithms, and promote scalability in simplifying them, in view of the great demand in the courts. In this context, this article presents an evaluation of four methods of state of the art in text simplification, evaluated according to readability metrics, to improve the quality of existing texts in the judicial summaries, dataset containing 100 summaries of the Federal Regional Court of the 5th Region (TRF5) and another 100 of the Federal Supreme Court (STF). The methods MUSS(EN), MUSS(PT), Transformers and NMT + Attention were tested, and the results of the simplifications exceeded the FRE readability index of the original texts, making them more readable.

A comunicação aumentada e alternativa é uma tecnologia assistiva que permite a pessoas não-verbais expressarem suas vontades e seus pensamentos. Porém, tipicamente, crianças não-verbais têm pranchas de comunicação preparadas para elas por adultos responsáveis, o que limita as suas possibilidades de expressão a partir da visão de uma outra pessoa. Em contextos lúdicos, que são fundamentais para o desenvolvimento infantil, essas limitações tornam-se mais gritantes e impeditivas. A partir de cenários de participação de crianças neurodivergentes e não-verbais em brincadeiras, apresentamos soluções que integram técnicas de IA gerativa e outros recursos a um aplicativo de comunicação alternativa, visando combater o capacitismo e a inclusão efetiva dessas crianças em brincadeiras, contribuindo para sua interação social e desenvolvimento cognitivo e emocional.

In the last decade, the study of pharmacological networks has received a lot of attention, given its relevance to the drug discovery process. Many different approaches for predicting biological interactions have been proposed, especially in the area of multiple kernel learning (MKL). Such methods comprise integrative approaches that can handle heterogeneous data sources in the form of kernels, but can suffer from the missing data problem. Techniques to handle missing values in the base kernel matrices can be used, usually based on simpler techniques, such as imputing zeroes, mean and median of the kernel matrix. In this work, techniques for handling missing values were evaluated in the context of bipartite networks. Our analyses showed that depending on the amount of missing data, k-NN and Singular Value Decomposition (SVD) techniques performed much better than the other techniques, bringing encouraging results, while zero-fill showed the worst performance in relation to all other evaluated methods.

Traditionally, single-domain recommender systems (SDRS) can suggest suitable products for users to alleviate information overload. Nonetheless, cross-domain recommender systems (CDRS) have enhanced SDRS by accomplishing specific objectives, such as improving precision and diversity and solving cold-start and sparsity issues. Rather than considering each domain separately, CDRS uses information gathered from a particular domain (e.g., music) to enhance recommendations for another domain (e.g., films). Context-aware Recommender System (CARS) focuses on optimizing the quality of suggestions, which are more appropriate for users depending on their context. Integrating these techniques is helpful for many cases where knowledge from several sources can be used to enhance recommendations and where relevant contextual information is considered. This work describes the main challenges and solutions of the state-of-the-art in Cross-Domain Context-Aware Recommender Systems (CD-CARS), taking into account the abundance of data on different domains and the systematic adoption of contextual data. CD-CARS have shown efficient methods to tackle realistic recommendation scenarios, preserving the benefits of CDRS (regarding cold-start and sparsity issues) and CARS (assuming accuracy). Therefore, CD-CARS may direct future research to recommender systems that use contextual information from multiple domains in a systematic way.

In recent years, the adoption of deep Convolutional Neural Networks (CNNs) has stood out in solving computer vision tasks, such as image classification. Researchers have proposed several architectures with varying sizes, complexities, and an increasing number of trainable parameters. For this reason, finding an optimized configuration and architecture with reduced complexity and high performance has become a very difficult task, since these configurations are totally dependent on the target classification problem and mostly depend on the optimization of a specialist in the area. To assist in the search for these optimal configurations, this work proposes the use of a multi-objective grammatical evolution framework, composed of a multi-objective search engine, a new context-free grammar responsible for creating the problem search space and a process mapping of individuals. Such a framework automatically generates and optimizes CNNs for a given image classification problem, without the need for human intervention from an expert. The framework navigates the search space using two objective functions seeking to maximize two metrics: accuracy and F1-score. The proposal was validated in the CIFAR-10, CIFAR-100, MNIST, KMNIST and EuroSAT datasets and the results show that the proposed method is able to generate simpler networks, but that statistically outperform (more complex) state-of-the-art CNNs in both metrics considered in the study.

Automated essay scoring (AES) is the task of automatically assigning scores (i.e., grades) to written texts. Although AES has been widely studied in the literature (e.g., informational and argumentative essays), specific types of texts still need more attention. Narrative essays are characterized by texts describing personal experiences and stories, either real or fictional. In this work, we describe a study on scoring student essays written in Portuguese under the aspect of Formal Register, which evaluates aspects related to the use of Brazilian Portuguese formal grammar and proficiency. The dataset created in this study provides a rich corpus of narrative essays produced in the context of a motivational situation, with a diverse set of language proficiency levels annotated by two professional graders. Different machine learning algorithms were evaluated using a diverse set of handcrafted linguistic features, and their results were compared against manual scores by the two human annotators. The results of the proposed analysis demonstrated that the AES model proposed achieved an equivalent agreement to that of the two human annotators.

Traditionally, single-domain recommender systems (SDRS) can suggest suitable products for users to alleviate information overload. Nonetheless, cross-domain recommender systems (CDRS) have enhanced SDRS by accomplishing specific objectives, such as improving precision and diversity and solving cold-start and sparsity issues. Rather than considering each domain separately, CDRS uses information gathered from a particular domain (e.g., music) to enhance recommendations for another domain (e.g., films). Context-aware Recommender System (CARS) focuses on optimizing the quality of suggestions, which are more appropriate for users depending on their context. Integrating these techniques is helpful for many cases where knowledge from several sources can be used to enhance recommendations and where relevant contextual information is considered. This work describes the main challenges and solutions of the state-of-the-art in Cross-Domain Context-Aware Recommender Systems (CD-CARS), taking into account the abundance of data on different domains and the systematic adoption of contextual data. CD-CARS have shown efficient methods to tackle realistic recommendation scenarios, preserving the benefits of CDRS (regarding cold-start and sparsity issues) and CARS (assuming accuracy). Therefore, CD-CARS may direct future research to recommender systems that use contextual information from multiple domains in a systematic way.

O Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira (INEP) disponibiliza o Censo da Educação Básica, o maior levantamento anual de dados sobre a educação brasileira. Os dados são disponibilizados anualmente e com cerca de 370 colunas e pouco mais de 230 mil registros por ano. Este trabalho apresenta o processo que foi utilizado para criar um conjunto de dados que unificasse os anos de 2010-2021 e o disponibilizasse de forma a garantir boas práticas de disponibilização de dados na web. Foi gerado um conjunto de dados abrangendo todos os anos mencionados, posteriormente dividido em subconjuntos dada a natureza dos dados apresentados.

Roles that learners assume during online discussions are an important aspect of educational experience. The roles can be assigned to learners and/or can spontaneously emerge through student-student interaction. While existing research proposed several approaches for analytics of emerging roles, there is limited research in analytic methods that can i) automatically detect emerging roles that can be interpreted in terms of higher-order constructs of collaboration; ii) analyse the extent to which students complied to scripted roles and how emerging roles compare to scripted ones; and iii) track progression of roles in social knowledge progression over time. To address these gaps in the literature, this paper propose a network-analytic approach that combines techniques of cluster analysis and epistemic network analysis. The method was validated in an empirical study discovered emerging roles that were found meaningful in terms of social and cognitive dimensions of the well-known model of communities of inquiry. The study also revealed similarities and differences between emerging and script roles played by learners and identified different progression trajectories in social knowledge construction between emerging and scripted roles. The proposed analytic approach and the study results have implications that can inform teaching practice and development techniques for collaboration analytics.

Neste estudo busca-se solucionar a falta de dados de estimativas populacionais segmentadas por município e idade, no período de 2014 a 2020 para todos os municípios do Brasil, através da criação de um Dataset que fornece estes dados de forma estruturada e enriquecida com características para facilitar seu reuso, partindo de dados oficiais como do IBGE e do Ministério da Saúde e processados por uma metodologia já aprovada por um órgão de Estado. Além da implantação da metodologia para geração do Dataset, também são discutidas oportunidades de melhoria no método de processamento, direcionando assim futuros estudos de desagregação populacional considerando as particularidades dos conjuntos de dados dos órgãos de Estado no Brasil.

Database processada dos anos de 2010 a 2021 do censo da educação básica fornecido pelo INEP. Os arquivos estão separados em bases diferentes e podem ser unidos utilizando o ano, código da instituição escolar e o código do município. Também é disponibilizado o dicionário de dados por meio de aquivos xlsx e ods. Os arquivos de colunas removidas e processadas estão disponíveis no formato xlsx. O dados originais que foram processados podem ser encontrados separadamente por ano na pagina do INEP. https://www.gov.br/inep/pt-br/acesso-a-informacao/dados-abertos/microdados/censo-escolar

Os chatbots são ferramentas que utilizam inteligência artificial para simular uma conversação humana. Eles podem ser utilizados para diferentes aplicações, como fornecer atendimento dentro de um e-commerce, preenchimento de formulários ou responder a FAQs (Frequently Asked Questions). Este trabalho propõe o desenvolvimento de um chatbot para auxiliar os estudantes de uma universidade pública brasileira na busca de informações relacionadas a dúvidas em disciplinas, processos administrativos, e dúvidas gerais sobre seu curso ou universidade. O sistema desenvolvido consegue entregar uma precisão alta na classificação da intenção da pergunta e ter resposta ao usuário em uma ampla margem de tópicos diferentes.

Domestic violence has increased globally as the COVID-19 pandemic combines with economic and social stresses. Some works have used traditional feature extractors to identify body positions to detect physical violence. Besides, the use of Machine Learning is limited by the trade-off between collecting more data while keeping users' privacy. Federated Learning (FL) is a technique that allows the creation of client-server networks, in which anonymized training data can be uploaded to a central model, responsible for aggregating and keeping the model up to date, and then distributing the updated model to the clients' nodes. This paper proposed an FL approach to the violence detection problem in video. The framework was evaluated on AIRTLab Dataset, in which frames were extracted. It used pretrained Convolutional Neural Networks (CNN) to address the image classification problem. Inception v3, MobileNet v2, ResNet152 v2, and VGG-16 architectures were evaluated, with the MobileNet architecture presenting the best performance, in terms of accuracy (99.4%), with a loss of 0.5% when compared to the non-FL setting.