Search

Decision Trees have remained a popular machine learning method for tabular datasets, mainly due to their interpretability. However, they lack the expressiveness needed to handle highly nonlinear or unstructured datasets. Motivated by recent advances in tree-based machine learning (ML) techniques and first-order optimization methods, we introduce Generalized Soft Trees (GSTs), which extend soft decision trees (STs) and are capable of processing images directly. We demonstrate their advantages with respect to tractability, performance, and interpretability. We develop a tractable approach to growing GSTs, given by the DeepTree algorithm, which, in addition to new regularization terms, produces high-quality models with far fewer nodes and greater interpretability than traditional soft trees. We test the performance of our GSTs on benchmark tabular and image datasets, including MIMIC-IV, MNIST, Fashion MNIST, CIFAR-10 and Celeb-A. We show that our approach outperforms other popular tree methods (CART, Random Forests, XGBoost) in almost all of the datasets, with Convolutional Trees having a significant edge in the hardest CIFAR-10 and Fashion MNIST datasets. Finally, we explore the interpretability of our GSTs and find that even the most complex GSTs are considerably more interpretable than deep neural networks. Overall, our approach of Generalized Soft Trees provides a tractable method that is high-performing on (un)structured datasets and preserves interpretability more than traditional deep learning methods., Comment: Submitted to Machine Learning. Authors are listed in alphabetical order

This article presents the OpenCitations Index, a collection of open citation data maintained by OpenCitations, an independent, not-for-profit infrastructure organisation for open scholarship dedicated to publishing open bibliographic and citation data using Semantic Web and Linked Open Data technologies. The collection involves citation data harvested from multiple sources. To address the possibility of different sources providing citation data for bibliographic entities represented with different identifiers, therefore potentially representing same citation, a deduplication mechanism has been implemented. This ensures that citations integrated into OpenCitations Index are accurately identified uniquely, even when different identifiers are used. This mechanism follows a specific workflow, which encompasses a preprocessing of the original source data, a management of the provided bibliographic metadata, and the generation of new citation data to be integrated into the OpenCitations Index. The process relies on another data collection: OpenCitations Meta, and on the use of a new globally persistent identifier, namely OMID (OpenCitations Meta Identifier). As of July 2024, OpenCitations Index stores over 2 billion unique citation links, harvest from Crossref, the National Institute of Heath Open Citation Collection (NIH-OCC), DataCite, OpenAIRE, and the Japan Link Center (JaLC). OpenCitations Index can be systematically accessed and queried through several services, including SPARQL endpoint, REST APIs, and web interfaces. Additionally, dataset dumps are available for free download and reuse (under CC0 waiver) in various formats (CSV, N-Triples, and Scholix), including provenance and change tracking information.

Identifying the reason for which an author cites another work is essential to understand the nature of scientific contributions and to assess their impact. Citations are one of the pillars of scholarly communication and most metrics employed to analyze these conceptual links are based on quantitative observations. Behind the act of referencing another scholarly work there is a whole world of meanings that needs to be proficiently and effectively revealed. This study emphasizes the importance of trustfully classifying citation intents to provide more comprehensive and insightful analyses in research assessment. We address this task by presenting a study utilizing advanced Ensemble Strategies for Citation Intent Classification (CIC) incorporating Language Models (LMs) and employing Explainable AI (XAI) techniques to enhance the interpretability and trustworthiness of models' predictions. Our approach involves two ensemble classifiers that utilize fine-tuned SciBERT and XLNet LMs as baselines. We further demonstrate the critical role of section titles as a feature in improving models' performances. The study also introduces a web application developed with Flask and currently available at http://137.204.64.4:81/cic/classifier, aimed at classifying citation intents. One of our models sets as a new state-of-the-art (SOTA) with an 89.46% Macro-F1 score on the SciCite benchmark. The integration of XAI techniques provides insights into the decision-making processes, highlighting the contributions of individual words for level-0 classifications, and of individual models for the metaclassification. The findings suggest that the inclusion of section titles significantly enhances classification performances in the CIC task. Our contributions provide useful insights for developing more robust datasets and methodologies, thus fostering a deeper understanding of scholarly communication., Comment: 42 pages, 14 figures, 1 table, submitted to Scientometrics Journal

In this article we analyse 3D models of cultural heritage with the aim of answering three main questions: what processes can be put in place to create a FAIR-by-design digital twin of a temporary exhibition? What are the main challenges in applying FAIR principles to 3D data in cultural heritage studies and how are they different from other types of data (e.g. images) from a data management perspective? We begin with a comprehensive literature review touching on: FAIR principles applied to cultural heritage data; representation models; both Object Provenance Information (OPI) and Metadata Record Provenance Information (MRPI), respectively meant as, on the one hand, the detailed history and origin of an object, and - on the other hand - the detailed history and origin of the metadata itself, which describes the primary object (whether physical or digital); 3D models as cultural heritage research data and their creation, selection, publication, archival and preservation. We then describe the process of creating the Aldrovandi Digital Twin, by collecting, storing and modelling data about cultural heritage objects and processes. We detail the many steps from the acquisition of the Digital Cultural Heritage Objects (DCHO), through to the upload of the optimised DCHO onto a web-based framework (ATON), with a focus on open technologies and standards for interoperability and preservation. Using the FAIR Principles for Heritage Library, Archive and Museum Collections as a framework, we look in detail at how the Digital Twin implements FAIR principles at the object and metadata level. We then describe the main challenges we encountered and we summarise what seem to be the peculiarities of 3D cultural heritage data and the possible directions for further research in this field.

As a multitude of capable machine learning (ML) models become widely available in forms such as open-source software and public APIs, central questions remain regarding their use in real-world applications, especially in high-stakes decision-making. Is there always one best model that should be used? When are the models likely to be error-prone? Should a black-box or interpretable model be used? In this work, we develop a prescriptive methodology to address these key questions, introducing a tree-based approach, Optimal Predictive-Policy Trees (OP2T), that yields interpretable policies for adaptively selecting a predictive model or ensemble, along with a parameterized option to reject making a prediction. We base our methods on learning globally optimized prescriptive trees. Our approach enables interpretable and adaptive model selection and rejection while only assuming access to model outputs. By learning policies over different feature spaces, including the model outputs, our approach works with both structured and unstructured datasets. We evaluate our approach on real-world datasets, including regression and classification tasks with both structured and unstructured data. We demonstrate that our approach provides both strong performance against baseline methods while yielding insights that help answer critical questions about which models to use, and when., Comment: Submitted to JMLR on 5/30/2024

The acquisition of physical artifacts not only involves transferring existing information into the digital ecosystem but also generates information as a process itself, underscoring the importance of meticulous management of FAIR data and metadata. In addition, the diversity of objects within the cultural heritage domain is reflected in a multitude of descriptive models. The digitization process expands the opportunities for exchange and joint utilization, granted that the descriptive schemas are made interoperable in advance. To achieve this goal, we propose a replicable workflow for metadata schema crosswalks that facilitates the preservation and accessibility of cultural heritage in the digital ecosystem. This work presents a methodology for metadata generation and management in the case study of the digital twin of the temporary exhibition "The Other Renaissance - Ulisse Aldrovandi and the Wonders of the World". The workflow delineates a systematic, step-by-step transformation of tabular data into RDF format, to enhance Linked Open Data. The methodology adopts the RDF Mapping Language (RML) technology for converting data to RDF with a human contribution involvement. This last aspect entails an interaction between digital humanists and domain experts through surveys leading to the abstraction and reformulation of domain-specific knowledge, to be exploited in the process of formalizing and converting information., Comment: Submitted to AIUCD conference 2024 1 figure 8 pages

As a result of the proliferation of 3D digitisation in the context of cultural heritage projects, digital assets and digitisation processes - being considered as proper research objects - must prioritise adherence to FAIR principles. Existing standards and ontologies, such as CIDOC CRM, play a crucial role in this regard, but they are often over-engineered for the need of a particular application context, thus making their understanding and adoption difficult. Application profiles of a given standard - defined as sets of ontological entities drawn from one or more semantic artefacts for a particular context or application - are usually proposed as tools for promoting interoperability and reuse while being tied entirely to the particular application context they refer to. In this paper, we present an adaptation and application of an ontology development methodology, i.e. SAMOD, to guide the creation of robust, semantically sound application profiles of large standard models. Using an existing pilot study we have developed in a project dedicated to leveraging virtual technologies to preserve and valorise cultural heritage, we introduce an application profile named CHAD-AP, that we have developed following our customised version of SAMOD. We reflect on the use of SAMOD and similar ontology development methodologies for this purpose, highlighting its strengths and current limitations, future developments, and possible adoption in other similar projects.

Remaining a dominant force in Internet traffic, video streaming captivates end users, service providers, and researchers. This paper takes a pragmatic approach to reviewing recent advances in the field by focusing on the prevalent streaming paradigm that involves delivering long-form two-dimensional videos over the best-effort Internet with client-side adaptive bitrate (ABR) algorithms and assistance from content delivery networks (CDNs). To enhance accessibility, we supplement the survey with tutorial material. Unlike existing surveys that offer fragmented views, our work provides a holistic perspective on the entire end-to-end streaming pipeline, from video capture by a camera-equipped device to playback by the end user. Our novel perspective covers the ingestion, processing, and distribution stages of the pipeline and addresses key challenges such as video compression, upload, transcoding, ABR algorithms, CDN support, and quality of experience. We review over 200 papers and classify streaming designs by their problem-solving methodology, whether based on intuition (simple heuristics), theory (formal optimization), or machine learning (generalizable data patterns). The survey further refines these methodology-based categories and characterizes each design by additional traits such as compatible codecs and use of super resolution. We connect the reviewed research to real-world applications by discussing the practices of commercial streaming platforms. Finally, the survey highlights prominent current trends and outlines future directions in video streaming.

One of the main goals of Open Science is to make research more reproducible. There is no consensus, however, on what exactly "reproducibility" is, as opposed for example to "replicability", and how it applies to different research fields. After a short review of the literature on reproducibility/replicability with a focus on the humanities, we describe how the creation of the digital twin of the temporary exhibition "The Other Renaissance" has been documented throughout, with different methods, but with constant attention to research transparency, openness and accountability. A careful documentation of the study design, data collection and analysis techniques helps reflect and make all possible influencing factors explicit, and is a fundamental tool for reliability and rigour and for opening the "black box" of research., Comment: Accepted to the AIUCD2024 Conference: https://aiucd2024.unict.it/ - will be published in conference proceedings

HERITRACE is a semantic data management system tailored for the GLAM sector. It is engineered to streamline data curation for non-technical users while also offering an efficient administrative interface for technical staff. The paper compares HERITRACE with other established platforms such as OmekaS, Semantic MediaWiki, Research Space, and CLEF, emphasizing its advantages in user friendliness, provenance management, change tracking, customization capabilities, and data integration. The system leverages SHACL for data modeling and employs the OpenCitations Data Model (OCDM) for provenance and change tracking, ensuring a harmonious blend of advanced technical features and user accessibility. Future developments include the integration of a robust authentication system and the expansion of data compatibility via the RDF Mapping Language (RML), enhancing HERITRACE's utility in digital heritage management., Comment: 5 pages, 1 figure, submitted to AIUCD 2024

This study describes the methodology and analyses the results of the process of mapping entities between two large open bibliographic metadata collections, OpenCitations Meta and OpenAlex. The primary objective of this mapping is to integrate OpenAlex internal identifiers into the existing metadata of bibliographic resources in OpenCitations Meta, thereby interlinking and aligning these collections. Furthermore, analysing the output of the mapping provides a unique perspective on the consistency and accuracy of bibliographic metadata, offering a valuable tool for identifying potential inconsistencies in the processed data.

As per the objectives of Project CHANGES, particularly its thematic sub-project on the use of virtual technologies for museums and art collections, our goal was to obtain a digital twin of the temporary exhibition on Ulisse Aldrovandi called "The Other Renaissance", and make it accessible to users online. After a preliminary study of the exhibition, focussing on acquisition constraints and related solutions, we proceeded with the digital twin creation by acquiring, processing, modelling, optimising, exporting, and metadating the exhibition. We made hybrid use of two acquisition techniques to create new digital cultural heritage objects and environments, and we used open technologies, formats, and protocols to make available the final digital product. Here, we describe the process of collecting and curating bibliographical exhibition (meta)data and the beginning of the digital twin creation to foster its findability, accessibility, interoperability, and reusability. The creation of the digital twin is currently ongoing.

The aim of this work is to understand the retraction phenomenon in the arts and humanities domain through an analysis of the retraction notices: formal documents stating and describing the retraction of a particular publication. The retractions and the corresponding notices are identified using the data provided by Retraction Watch. Our methodology for the analysis combines a metadata analysis and a content analysis (mainly performed using a topic modeling process) of the retraction notices. Considering 343 cases of retraction, we found that many retraction notices are neither identifiable nor findable. In addition, these were not always separated from the original papers, introducing ambiguity in understanding how these notices were perceived by the community (i.e., cited). Also, we noticed that there is no systematic way to write a retraction notice. Indeed, some retraction notices presented a complete discussion of the reasons for retraction, while others tended to be more direct and succinct. We have also reported many notices having similar text while addressing different retractions. We think a further study with a larger collection should be done using the same methodology to confirm and investigate our findings further.

The paper introduces a tool prototype that combines SHACL's capabilities with ad-hoc validation functions to create a controlled and user-friendly form interface for producing valid RDF data. The proposed tool is developed within the context of the OpenCitations Data Model (OCDM) use case. The paper discusses the current status of the tool, outlines the future steps required for achieving full functionality, and explores the potential applications and benefits of the tool.

OpenCitations Meta is a new database that contains bibliographic metadata of scholarly publications involved in citations indexed by the OpenCitations infrastructure. It adheres to Open Science principles and provides data under a CC0 license for maximum reuse. The data can be accessed through a SPARQL endpoint, REST APIs, and dumps. OpenCitations Meta serves three important purposes. Firstly, it enables disambiguation of citations between publications described using different identifiers from various sources. For example, it can link publications identified by DOIs in Crossref and PMIDs in PubMed. Secondly, it assigns new globally persistent identifiers (PIDs), known as OpenCitations Meta Identifiers (OMIDs), to bibliographic resources without existing external persistent identifiers like DOIs. Lastly, by hosting the bibliographic metadata internally, OpenCitations Meta improves the speed of metadata retrieval for citing and cited documents. The database is populated through automated data curation, including deduplication, error correction, and metadata enrichment. The data is stored in RDF format following the OpenCitations Data Model, and changes and provenance information are tracked. OpenCitations Meta and its production. OpenCitations Meta currently incorporates data from Crossref, DataCite, and the NIH Open Citation Collection. In terms of semantic publishing datasets, it is currently the first in data volume., Comment: 26 pages, 7 figures

In this paper we study the algebraic properties of a new integrable differential-difference equation. This equation can be seen as a deformation of the modified Narita-Itoh-Bogoyavlensky equation and has the Kaup-Kupershmidt equation in its continuous limit. Using its Lax representation we explicitly construct a recursion operator for this equation and prove that it is a Nijenhuis operator. Moreover, we present the bi-Hamiltonian structures for this new equation., Comment: 11 pages

In the realm of Digital Humanities, the management of cultural heritage metadata is pivotal for ensuring data trustworthiness. Provenance information - contextual metadata detailing the origin and history of data - plays a crucial role in this process. However, tracking provenance and changes in metadata using the Resource Description Framework (RDF) presents significant challenges due to the limitations of foundational Semantic Web technologies. This article offers a comprehensive review of existing models and approaches for representing provenance and tracking changes in RDF, with a specific focus on cultural heritage metadata. It examines W3C standard proposals such as RDF Reification and n-ary relations, along with various alternative systems. Through an in-depth analysis, the study identifies Named Graphs, RDF*, the Provenance Ontology (PROV-O), Dublin Core (DC), Conjectural Graphs, and the OpenCitations Data Model (OCDM) as the most effective solutions. These models are evaluated based on their compliance with RDF standards, scalability, and applicability across different domains. The findings underscore the importance of selecting the appropriate model to ensure robust and reliable management of provenance in RDF datasets, thereby contributing to the ongoing discourse on provenance representation in the Digital Humanities., Comment: 23 pages, 4 figures, submitted to Digital Scholarship in the Humanities

This work presents a maturity model for assessing catalogues of semantic artefacts, one of the keystones that permit semantic interoperability of systems. We defined the dimensions and related features to include in the maturity model by analysing the current literature and existing catalogues of semantic artefacts provided by experts. In addition, we assessed 26 different catalogues to demonstrate the effectiveness of the maturity model, which includes 12 different dimensions (Metadata, Openness, Quality, Availability, Statistics, PID, Governance, Community, Sustainability, Technology, Transparency, and Assessment) and 43 related features (or sub-criteria) associated with these dimensions. Such a maturity model is one of the first attempts to provide recommendations for governance and processes for preserving and maintaining semantic artefacts and helps assess/address interoperability challenges.