Currently, governments and many other institutions have been making significant efforts to promote financial literacy. However, a considerable portion of the population still lacks basic financial knowledge, highlighting the need for updated strategies to enhance financial education. In today's digital world - where people often search for quick and convenient solutions - the development of a reliable and intelligent chatbot to answer questions related to financial concepts and decision-making can be very beneficial. This paper proposes the implementation of an automated web scraper to extract content from a trustworthy financial education website with plenty of useful concepts about finances, using this collected data to develop a chatbot which provides accurate and helpful responses to users. The solution is built using technologies such as Streamlit, Langchain, and OpenAI.
In this paper, we introduce elements for MoSCO, a framework for building hybrid metaheuristic-based solvers from a collection of reusable base components. The framework is implemented in Julia and provides a modular architecture for composing solvers through a pipeline-based approach. The modular design of MoSCO supports the creation of reusable components and adaptable solver strategies for various Constraint Satisfaction Problems (CSPs) and Constraint Optimization Problems (COPs). We validate MoSCO's utility through practical examples, demonstrating its effectiveness in reconstructing established metaheuristics and enabling the creation of novel solver configurations. This work lays the foundation for future developments in automated solver construction and parameter optimization.
This study analyzes the social media discourse of leading figures from Portugal's far right party CHEGA, examining 10,323 posts on X (formerly Twitter) published between late 2019 and mid-2024. Using BERTopic, 59 latent topics clustered into two main discursive dynamics were found: (1) ideological and public, and (2) party, electoral and parliamentary related. Within the first dynamic, we conducted a focused sub-analysis of themes related with identity, immigration and security narratives - topics that display posting peaks around electoral cycles, suggesting the strategic use of emotionally charged, identitarian frames for political mobilization. The model exhibits strong topic coherence and lexical diversity, indicating its robustness in extracting thematic structures from politically polarized microtexts. Nevertheless, our findings are constrained by source, the absence of interaction metrics, and the unmet need to link online discourse to offline events. This study demonstrates how computational topic modeling can reveal strategic communication patterns in far-right political discourse and underscores the need for cross-platform and interaction-level research to assess broader societal impact.
This study examines the intersection of sociodemographic characteristics, linguistic features, and writing placement outcomes at a community college in the United States of America. It focuses on 210 anonymized writing samples from native English speakers (L1) that were automatically classified by ACCUPLACER and independently assessed by two trained raters. Disparities across gender and race using 40 top-ranked linguistic features selected from Coh-Metrix, CTAP, and Developmental Education-Specific (DES) sets were analyzed. Three statistical tests were used: one-way ANOVA, Tukey's HSD, and Chi-square. ANOVA results showed racial differences in nine linguistic features, especially those tied to syntactic complexity, discourse markers, and lexical precision. Gender differences were more limited, with only one feature reaching significance (Positive Connectives, p = 0.007). Tukey's HSD pairwise tests showed no significant gender group variation but revealed sensitivity in DES features when comparing racial groups. Chi-square analysis indicated no significant association between gender and placement outcomes but suggested a possible link between race and human-assigned levels (chi(2) = 9.588, p = 0.048). These findings suggest that while automated systems assess general writing skills, human-devised linguistic features and demographic insights can support more equitable placement practices for all students entering college-level programs.
Semantic parsing serves as a crucial interface between natural language and formal meaning representations, enabling computational systems to capture the underlying semantic structure of linguistic expressions. This paper addresses a relatively understudied area in both linguistic theory and natural language processing: the semantic representation of adverbs. We conduct a comparative analysis of annotation guidelines and practices across two semantic representation frameworks: Lexicalized Meaning Representation (LMR), applied to the European Portuguese edition of the novella "O Principezinho" by Antoine de Saint-Exupery (1943); and Abstract Meaning Representation (AMR), applied to the Brazilian Portuguese edition, "O Pequeno Principe". The study reveals significant limitations in AMR's handling of adverbial constructions, particularly when assessed against contemporary syntactic-semantic advances in linguistic theory. Furthermore, it highlights the theoretical and practical challenges that LMR continues to face in this domain.
We propose a domain-specific language to simplify the expression of Swarm Intelligence algorithms. These algorithms are typically introduced through metaphors, requiring practitioners to manually translate them into low-level implementations.This process can obscure intent and hinder reproducibility. The proposed DSL bridges this gap by capturing algorithmic behavior at a higher level of abstraction. We demonstrate its expressiveness in a few lines of code and evaluate its feasibility through a reference implementation. A discussion is presented that includes empirical comparisons with traditional implementations and future directions of the proposed DSL.
In light of specific development needs, it is common to concurrently apply different technologies to build complex applications. Given that lowering risks, costs, and other negative factors, while improving their positive counterparts is paramount to a better development environment, it becomes relevant to find out what technologies work best for each intended purpose in a project. In order to reach these findings, it is necessary to analyse and study the technologies applied in these projects and how they interconnect and relate to each other. The theory behind Programming Cocktails (meaning the set of programming technologies - Ingredients - that are used to develop complex systems) can support these analysis. However, due to the sheer amount of data that is required to construct and analyse these Cocktails, it becomes unsustainable to manually obtain them. From the desire to accelerate this process comes the need for a tool that automates the data collection and its conversion into an appropriate format for analysis. As such, the project proposed in this paper revolves around the development of a web-scraping application that can generate Cocktail Identity Cards (CIC) from source code repositories hosted on GitHub. Said CICs contain the Ingredients (programming languages, libraries and frameworks) used in the corresponding GitHub repository and follow the ontology previously established in a larger research project to model each Programming Cocktail. This paper presents a survey of current Source Version Control Systems (SVCSs) and web-scrapping technologies, an overview of Programming Cocktails and its current foundations, and the design of a tool that can automate the gathering of CICs from GitHub repositories.
As academic professionals, we frequently need to create different versions of our CVs for project applications, career evaluations, or competitions. These versions may be chronologically structured or skill-oriented, covering specific periods and written in various languages. Even when using a LaTeX document as a base, numerous modifications are required each time an updated CV version is requested for a specific purpose. The primary objective of the project reported in this paper is to design and implement a web-based system (CVTool) that simplifies the management of LaTeX CV content while ensuring flexibility. The CVTool is built on a domain-specific language that enables the creation of various filters, allowing for the automatic content adjustment while preserving the original format. Information is extracted from the LaTeX document, and users can specify the sections, dates, skills they want to highlight, and the language in which the CV should be generated. Since the approach relies on an internal data representation derived from the original LaTeX document, it offers users the flexibility to manage content efficiently and extract the necessary information with ease.
In recent years, more attention has been paid to the security of the software supply chain (SSC). While at first SSC was just seen as the dependency of other libraries, today SSC is broader, considering all the environment where a software application is developed, from the actors, hardware and auxiliary tools. This work focuses on a specific part of software supply chain security: the tools used for continuous integration and continuous deployment (CI/CD) and their vulnerabilities and risks. These tools are widely used by organizations to accelerate their software development, testing, and delivery, making any security issue present in these tools problematic for the organization. This is especially true given that most tools are open-source, making these tools the primary targets for exploits. We will present a quick introduction to SSCS and CI/CD and provide a practical solution to detect risks and vulnerabilities in CI/CD tools, emphasizing the modular approach, allowing the system to easily scale to detect new risks and vulnerabilities, as well as to support new CI/CD tools.
Source-to-source transpilation plays a pivotal role in modern software engineering by enabling code migration, feature adoption, and cross-language interoperability without sacrificing semantic integrity. The contributions discussed in this paper can be split into two. The first is a comprehensive literature review that aims at defining what transpilers are, traces their historical evolution from early Fortran/COBOL preprocessors to more recent tools like Babel and TypeScript, and examines key parsing methodologies, AST representations, and transformation strategies. The second is an experimental investigation which assesses several popular transpilers - selected by GitHub popularity and unique language-pair capabilities, when applied to an equivalent code snippet designed to sum even numbers and identify the maximum element. The metrics evaluated were the execution time, CPU, memory consumption, output accuracy and usability.
Logic programs, more specifically, answer-set programs, can be annotated with probabilities on facts to express uncertainty. We address the problem of propagating weight annotations on facts (e.g. probabilities) of an answer-set program to its stable models, and from there to events (defined as sets of atoms) in a dataset over the program's domain. We propose a novel approach which is algebraic in the sense that it relies on an equivalence relation over the set of events. Uncertainty is then described as polynomial expressions over variables. We propagate the weight function in the space of models and events, rather than doing so within the syntax of the program. As evidence that our approach is sound, we show that certain facts behave as expected. Our approach allows us to investigate weight annotated programs and to determine how suitable a given one is for modeling a given dataset containing events. It's core is illustrated by a running example and the encoding of a Bayesian network.
The difficulties in teaching and learning computer programming remain a pressing issue to this day. Several studies and tools have been developed over the years to tackle this challenge from many different points-of-view. One of the biggest tools an educator has to support him in a classroom is the progressive explanation of how a source code is constructed and what effects each of its parts has on the overall result. Attempts to translate this live-directed tool to an on-line experience is usually time-consuming and lacking in features. In order to tackle this concern, a tool to create piecewise source code writing demonstrations was developed - Stepwise Source. The main idea behind this application is to allow step-by-step explanation of a source code construction, along with any relevant annotations and automatically assessed challenges that an educator may add. By providing a dynamic platform for both students and lecturers, this software aims to improve the teaching and learning of computer programming, while trying to imitate the information flow of a live lecture, with the added benefit of student-directed pace of explanation. Through interactive guidance and automated assessment, this tool has the potential to foster a deeper understanding of computational principles and promote proficiency in programming skills.
Accurate text classification and placement remain challenges in U.S. higher education, with traditional automated systems like ACCUPLACER functioning as "black-box" models with limited assessment transparency. This study evaluates Large Language Models (LLMs) as complementary placement tools by comparing their classification performance against a human-rated gold standard and ACCUPLACER. A 450-essay corpus was classified using Claude, Gemini, GPT-3.5-turbo, and GPT-4o across four prompting strategies: Zero-shot, Few-shot, Enhanced, and Enhanced+ (definitions with examples). Two classification approaches were tested: (i) a 1-step, 3 class classification task, distinguishing DevEd Level 1, DevEd Level 2, and College-level texts in one single run; and (ii) a 2-step classification task, first separating College vs. Non-College texts before further classifying Non-College texts into DevEd sublevels. The results show that structured prompt refinement improves the precision of LLMs' classification, with Claude Enhanced + achieving 62.22% precision (1 step) and Gemini Enhanced + reaching 69.33% (2 step), both surpassing ACCUPLACER (58.22%). Gemini and Claude also demonstrated strong correlation with human ratings, with Claude achieving the highest Pearson scores (rho = 0.75; 1-step, rho = 0.73; 2-step) vs. ACCUPLACER (rho = 0.67). While LLMs show promise for DevEd placement, their precision remains a work in progress, highlighting the need for further refinement and safeguards to ensure ethical and equitable placement.
In the context digital transformation, the necessity for secure and efficient virtual identity verification has become paramount. Traditional methods often fail to balance security, speed, and usability, leaving gaps in user authentication systems. This paper addresses the critical challenge of creating a virtual ID system that identifies a single profile with improved security, speed, and effectiveness. An innovative face recognition algorithm using dynamic content loading and deep learning techniques is proposed. The utilisation of OpenCV for face recognition and feature extraction, combined with advanced similarity calculation methods, the system achieves superior accuracy in profile authentication tasks. Extensive testing, including identical twin scenarios, demonstrates the robustness of the algorithm and its superiority over existing solutions such as Apple's Face ID. In 150 of the tests conducted with identical twins, the algorithm consistently achieved 100% recognition accuracy. This breakthrough in facial recognition technology promises to create a triple authentication system, which will solve the problem of false positives in terms of identifying and authenticating people. This paper integrates principles from Computer Intelligence and Chatbots, emphasizing the application of deep learning techniques in enhancing virtual identity verification systems. This research contributes to the broader discourse on improving authentication mechanisms in the digital age.
Organizations are facing some challenges in cybersecurity, due to the increasing of cyber threats, vulnerabilities, insufficient cybersecurity frameworks, and scarcity of proficient cybersecurity experts. The criticality of mitigating these challenges is underscored by the European Union's Network and Information Systems (NIS) Directive. This directive is instrumental in fostering a uniformly high level of cybersecurity throughout the EU, mandating that Member States implement robust national cybersecurity strategies and collaborate effectively in responding to cyber incidents. A possible solution is the implementation of a Security Operations Center (SOC). However, SOCs are not a one-size-fits-all solution and each organization has specific needs depending on their business domain. This task can be complex, and it can be simplified when organizations can identify in the initial stages the infrastructural challenges that may emerge when implementing a SOC. In this paper, we analyzed the main considerations that should be considered when using current frameworks reviewed in the literature. We identified the core operating models that are currently in use and being deployed, and which are the best practices when designing a SOC's infrastructure.
Semantic measure (SM) algorithms allow software to mimic the human ability of assessing the strength of the semantic relations between elements such as concepts, entities, words, or sentences. SM algorithms are typically evaluated by comparison against gold standard datasets built by human annotators. These datasets are composed of pairs of elements and an averaged numeric rating. Building such datasets usually requires asking human annotators to assign a numeric value to their perception of the strength of the semantic relation between two elements. Large language models (LLMs) have recently been successfully used to perform tasks which previously required human intervention, such as text summarization, essay writing, image description, image synthesis, question answering, and so on. In this paper, we present ongoing research on LLMs capabilities for semantic relations assessment. We queried several LLMs to rate the relationship of pairs of elements from existing semantic measures evaluation datasets, and measured the correlation between the results from the LLMs and gold standard datasets. Furthermore, we performed additional experiments to evaluate which other factors can influence LLMs performance in this task. We present and discuss the results obtained so far.
Enterprise Wi-Fi networks provide seamless connectivity for businesses and public entities. In educational settings, these networks are essential for delivering services to students. Eduroam is one of the most widely recognized university Wi-Fi enterprise networks. It uses 802.1x protocols and TLS tunnels to establish mutual authentication. Eduroam caters to a diverse population of students and faculty worldwide, managing a vast number of unregulated devices, some of which, require manual configuration to operate in a secure setting. This study aims to evaluate the security robustness of the 802.1x protocol, particularly within the Eduroam framework. A comprehensive study was conducted to examine compliance with institutional instructions for network configuration guidance. The results from 91 Portuguese institutions using Eduroam revealed that many lacked knowledge of Wi-Fi security. This lack of awareness is then passed on to users, resulting in vulnerable connections.
This document presents an initiative aimed at advancing parsing and language processing through the integration of Attribute Grammars (AG) into the Lark Compiler Generator, a flexible tool frequently used for these tasks. In order to achieve a successful integration of AGs into Lark, a study on the concept and example analysis of AG based specifications needed to be conducted. This provided an insight on their advantages in order to extend Lark with the ability to use AG, that will be presented in this article. With this project, named LarkAG, we aim at providing Lark with a conventional and well studied formalism to specify rigorously the static and dynamic semantics of programming languages, on top of its recognized syntactic analysis capabilities. Along the paper, LarkAG architecture, development and usage are also discussed. The DSL designed to provide a proper notation for attribute occurrences selection and for writing semantic rules and context conditions is also presented. This addition of AG support to Lark will enable the construction of compilers and language processors of greater reliability. It is not a major breakthrough since AGs based tools already exist and are well studied, but rather an extension that can bring great value to a modern framework such as Lark.
This study explores anomaly detection through unsupervised Machine Learning applied to banking systems' log records. The diversity in formatting and types of logs poses significant challenges for automating anomaly detection. We propose a workflow using Natural Language Processing (NLP) techniques for anomaly identification, which in further analysis can lead to identifying root causes of failures and vulnerabilities. We evaluate the performance of eight different models using Blue Gene/L log records. The most effective models were selected and subsequently validated with Microsoft Configuration Manager (MCM) logs collected from a financial institution, demonstrating their practical applicability in real-world scenarios. Experimental results highlighted the effectiveness of neural network models, specifically Self-Organizing Maps (SOM) and Autoencoders (AE), with F1-Scores of 0.86 and 0.80, respectively, when applied to MCM logs collected from the financial institution.
Editors can significantly influence user experience, as well as factors such as programming speed and the occurrence of syntactic errors. Given our experience in programming in the projectional editor MPS, which we believe offers numerous advantages, we decided to conduct an experiment involving students with limited programming experience. The aim was to determine if a novice programmer could program in projectional editor more effectively than in traditional text editor. A total of 83 first-year computer science students from our university participated in the experiment. The methodology encompassed the selection of an appropriate programming language for the experiment, customization of the language's editor, preliminary trials, the main experiment, and concluding evaluations of speed and error rates associated with both editors. Additionally, the study sought insights into whether students exhibited a preference for programming in a projectional editor or adhering to a classic text editor paradigm.
