Search

Generating synthetic data is a complex task that necessitates accurately replicating the statistical and mathematical properties of the original data elements. In sectors such as finance, utilizing and disseminating real data for research or model development can pose substantial privacy risks owing to the inclusion of sensitive information. Additionally, authentic data may be scarce, particularly in specialized domains where acquiring ample, varied, and high-quality data is difficult or costly. This scarcity or limited data availability can limit the training and testing of machine-learning models. In this paper, we address this challenge. In particular, our task is to synthesize a dataset with similar properties to an input dataset about the stock market. The input dataset is anonymized and consists of very few columns and rows, contains many inconsistencies, such as missing rows and duplicates, and its values are not normalized, scaled, or balanced. We explore the utilization of generative adversarial networks, a deep-learning technique, to generate synthetic data and evaluate its quality compared to the input stock dataset. Our innovation involves generating artificial datasets that mimic the statistical properties of the input elements without revealing complete information. For example, synthetic datasets can capture the distribution of stock prices, trading volumes, and market trends observed in the original dataset. The generated datasets cover a wider range of scenarios and variations, enabling researchers and practitioners to explore different market conditions and investment strategies. This diversity can enhance the robustness and generalization of machine-learning models. We evaluate our synthetic data in terms of the mean, similarities, and correlations.

The ongoing transformation of society influences and pushes educational institutions to innovate and adapt. This does not necessarily have positive repercussions on the improvement of educational models. New problems and needs arise that require a new type of research, reflection and action. The idea behind my contribution is anchored in decades of experience. This is exactly what I will try to explain without periphrasis: philosophy (and not just the philosophy of education) is no longer interesting today because it seems neither understandable nor useful. Among the various scholars interested in a new perspective, Ikeda creates a truly new, but exquisitely genuine approach. This approach starts from wonder, not from theory, from amazement, from emotion, from everything that is located in the person and it does not need to be translated into a distorting language.

Numerous methods and pipelines have recently emerged for the automatic extraction of knowledge graphs from documents such as scientific publications and patents. However, adapting these methods to incorporate alternative text sources like micro-blogging posts and news has proven challenging as they struggle to model open-domain entities and relations, typically found in these sources. In this paper, we propose an enhanced information extraction pipeline tailored to the extraction of a knowledge graph comprising open-domain entities from micro-blogging posts on social media platforms. Our pipeline leverages dependency parsing and classifies entity relations in an unsupervised manner through hierarchical clustering over word embeddings. We provide a use case on extracting semantic triples from a corpus of 100 thousand tweets about digital transformation and publicly release the generated knowledge graph. On the same dataset, we conduct two experimental evaluations, showing that the system produces triples with precision over 95% and outperforms similar pipelines of around 5% in terms of precision, while generating a comparatively higher number of triples.

In video analysis, collection and labeling of data can be time and resource-consuming. To solve the scarcity of data problems, synthetic data augmentation is a promising solution. In this paper, we present an approach to generate synthetic videos for action recognition using Unity, the popular game engine. The synthetic videos are generated with high variability in lighting, subjects’ models, backgrounds, animations, and camera positions. We use the generated data to augment a small dataset of subjects who are executing physical exercises for action recognition. We tested the augmented data on two state-of-the-art models for action classification and demonstrated the significant benefits of synthetic data augmentation for improving the performance of these models on small datasets in the context of video action recognition.

In today’s rapidly evolving labor market, the emergence of new roles and the decline of traditional ones have led to a complex landscape of job titles and skill requirements. This complexity often causes ambiguity and confusion, affecting both novices and experienced professionals. To address this, extensive international efforts have produced reference databases of jobs and skills, such as ESCO and O*NET. However, the challenge remains to make this information easily accessible and interpretable for users with varying levels of expertise. To address the challenge above, this paper introduces a Knowledge Plug-in for ChatGPT, designed to serve as an intuitive, user-friendly interface between workers and these authoritative databases. By harnessing the power of natural language processing (NLP), the plug-in enables a seamless question-answering experience, effectively masking the underlying complexity with a carefully engineered architecture. Furthermore, generative AI enhances the user experience by providing relevant information in domains extending beyond the traditional scope of employment. An initial user study demonstrates the plug-in’s effectiveness in improving the usability and accuracy of job-related queries. We detail the development, architecture, and validation of this innovative tool, highlighting its potential impact on the future of employment search and career development.

Machine and Deep Learning methods are widely adopted to predict corporate bankruptcy events for their effectiveness. Bankruptcy prediction is commonly modeled as a binary classification task over accounting data where the positive label is associated with companies with a high likelihood of bankruptcy and the negative label with a low risk of failure. Most of the models mainly focus on exploiting accounting, stock market data, and data augmentation to deal with the intrinsic unbalance of this task. More recently, financial reports such as the US SEC annual reports have been investigated for feature engineering to boost the accuracy of the classification task. However, these approaches only marginally leverage Natural Language Processing advanced techniques to improve the prediction, by usually only leveraging dictionary-based approaches and word frequencies for feature engineering. These fixed features suffer from concept drift over time leading to weaker predictive models by missing a data-driven architecture to extract text disclosures from financial reports to improve the task. This paper aims to fill the gap between the bankruptcy prediction domain and the recent advances in Natural Language Processing by proposing a Transformer-based architecture that combines: a) a text summarization module that extracts text disclosures from financial reports, over time, by leveraging the self-attention mechanism and learning which contents are more valuable for the prediction in a text communication; b) a multivariate time series modeling for accounting data that is aligned and optimized along with the text module. In this way, the architecture benefits from both data sources for the prediction and ensures continual model adaptation over time. We focused on public companies listed in the American stock market with a dataset including 6190 companies from 1999 to 2018. We have deeply analyzed the contribution of the two proposed modules, the accounting time series module (Accuracy 78%) and the text disclosures module (Accuracy 81%) to finally prove that a unique model that can leverage both data sources at the same time achieves better performance (Accuracy 87.5%). The architecture also outperforms the other baselines for Recall of default events (0.84) and for type II error (16.12).

The labor market is a dynamic and rapidly evolving environment. Job positions that require expertise in various sectors often lead candidates to question their suitability. Therefore, it is crucial to furnish them with relevant, accurate, and timely information. In this article, we introduce a knowledge plug-in for existing conversational agents designed to address specific queries related to the job market domain. Additionally, we propose an innovative method for dynamically creating diverse grammars and semantically associating users’ questions with a predefined list of domain questions. We present a novel scheme to effectively tackle question-answering tasks in settings with limited resources. Our architecture relies on question-understanding and response-generation modules, both powered by Transformers, the O*NET occupational database, and recommendation engines that suggest training materials. Furthermore, we conducted a user study based on the System Usability Scale (SUS) score, revealing that users highly appreciated the proposed tool. This sentiment was particularly evident when the tool was integrated with other artificial intelligence chatbots capable of handling general information. Simultaneously, our engine adeptly manages information within the investigated domain, providing precise responses and recommendations. This work addresses a critical gap in the delivery of employment information and paves the way for the development of diverse functionalities to assist both candidates and employers.

In recent years, the significance of Environmental, Social, and Governance criteria in assessing financial investments has grown significantly. This paper presents an AI-driven analysis of ESG concepts and their evolution from 1980 to 2022, with a specific focus on media sources from the United States and the United Kingdom. The primary data source utilized is the Dow Jones News Article dataset, providing a comprehensive and high-quality collection of news articles. The study introduces a novel technique for information extraction from news articles, involving the structuring of extracted data into a knowledge graph. The findings identified key trends associated with ESG aspects emerging in recent years. In the environmental dimension, we identified pronounced emphasis on climate change, renewable energy sources, and biodiversity conservation. Within the social aspect, the analysis pointed out the increasing significance of issues such as racism, gender identity, human rights as well all the increasing role of charities, and the ethical challenges of modern supply chains. Finally, in the governance domain, the findings emphasized issues related to corporate governance accountability, workplace ethics, and the conduct and remuneration of executives.

In an attempt to do a back analysis of the damages caused to a nineteenth century masonry structure due to the October 2009 flash flood/debris flow event in Scaletta Zanclea, the flood discharge hydrograph is reconstructed in the ungauged conditions. The hydrograph for the solid discharge is then estimated by scaling up the liquid volume to the estimated debris volume. The debris flow diffusion is simulated by solving the differential equations for a single-phase 2D flow employing triangular mesh elements, taking into account also the channelling of the flow through the buildings. The damage to the building is modelled, based on the maximum hydraulic actions caused by the debris flow, using 2D finite shell elements to model the building, boundary conditions provided by the openings, floor slab, orthogonal wall panels, and the foundation. The reconstruction of the event and the damages to the case-study building confirms the location of the damages induced by the event.

ICOM (International Council of Museums) envisions a role for the museum as an accessible and inclusive institution open to the public, which operates professionally and with the participation of communities. This vision requires research and practice to be fully implemented. The paper illustrates the participatory ideation and evaluation of new museum services that are inclusive and accessible for people who cannot visit the museum due to disability and age-related impairments. It illustrates the co-design process carried out within the EU project BeauCoup, which aims to make the cultural heritage accessible beyond the museum’s walls. This objective implies that the museum must play a key role in promoting inter-institutional partnerships between the cultural and the social sectors to engage marginalised communities. The paper describes the approach and the outcomes of the co-design process which involved various stakeholders and institutions, offering a reflection on the benefits and shortcomings of the approach.

book review

Almost all individuals with Down’s syndrome (DS) show the characteristic neuropathological features of Alzheimer’s disease (AD) by the age of 40, yet not every individual with DS experiences symptoms of AD later in life. Similar to neurotypical developing subjects, AD in people with DS lasts for a long preclinical phase in which biomarkers follow a predictable order of changes. Hence, a prolonged asymptomatic period precedes the onset of dementia, underscoring the importance of identifying new biomarkers for the early detection and monitoring of cognitive decline in individuals with DS. Blood-based biomarkers may offer an alternative non-invasive strategy for the detection of peripheral biological alterations paralleling nervous system pathology in an early phase of the AD continuum. In the last few years, a strong neurobiological link has been demonstrated between the deficit of transforming growth factor-β1 (TGF-β1) levels, an anti-inflammatory cytokine endowed with neuroprotective activity, and early pro-inflammatory processes in the AD brain. In this clinical prospective observational study, we found significant lower plasma TGF-β1 concentrations at the first neuropsychological evaluation (baseline = T0) both in young adult DS individuals (19–35 years) and older DS subjects without AD (35–60 years) compared to age- and sex-matched healthy controls. Interestingly, we found that the lower TGF-β1 plasma concentrations at T0 were strongly correlated with the following cognitive decline at 12 months. In addition, in young individuals with DS, we found, for the first time, a negative correlation between low TGF-β1 concentrations and high TNF-α plasma concentrations, a pro-inflammatory cytokine that is known to be associated with cognitive impairment in DS individuals with AD. Finally, adopting an ex vivo approach, we found that TGF-β1 concentrations were reduced in parallel both in the plasma and in the peripheral blood mononuclear cells (PBMCs) of DS subjects, and interestingly, therapeutic concentrations of fluoxetine (FLX) applied to cultured PBMCs (1 µM for 24 h) were able to rescue TGF-β1 concentrations in the culture media from DS PBMCs, suggesting that FLX, a selective serotonin reuptake inhibitor (SSRI) endowed with neuroprotective activity, might rescue TGF-β1 concentrations in DS subjects at higher risk to develop cognitive decline.