• Digital Humanities and Cultural Heritage
• Publications close
• Research data close
• Research software close
• Other research products close
• Research close
• NARCIS close

Research assessment reform is crucial for the social sustainability of research infrastructures (RIs): RIs can only thrive in the long term if the researchers who contribute to their development and growth receive academic credit for the kind of work they do in and around research infrastructures. To put it bluntly, research infrastructures have a vested interest in supporting the reform of research assessment. But, conversely, ongoing attempts to reform research assessment can also benefit from the work of research infrastructures because RIs have a great deal of experience creating and maintaining public services for producing, curating and harvesting both traditional and non-traditional academic outputs. The goal of this paper is to outline DARIAH’s position on the importance of research assessment reform for thematic RIs and the importance of thematic RIs for research assessment reform at the European level.

Past work in natural language processing interpretability focused mainly on popular classification tasks while largely overlooking generation settings, partly due to a lack of dedicated tools. In this work, we introduce Inseq, a Python library to democratize access to interpretability analyses of sequence generation models. Inseq enables intuitive and optimized extraction of models' internal information and feature importance scores for popular decoder-only and encoder-decoder Transformers architectures. We showcase its potential by adopting it to highlight gender biases in machine translation models and locate factual knowledge inside GPT-2. Thanks to its extensible interface supporting cutting-edge techniques such as contrastive feature attribution, Inseq can drive future advances in explainable natural language generation, centralizing good practices and enabling fair and reproducible model evaluations. Comment: ACL 2023 Demo Track. Library: https://github.com/inseq-team/inseq, Docs: https://inseq.readthedocs.io, v0.4

The goal of a speech-to-image transform is to produce a photo-realistic picture directly from a speech signal. Recently, various studies have focused on this task and have achieved promising performance. However, current speech-to-image approaches are based on a stacked modular framework that suffers from three vital issues: 1) Training separate networks is time-consuming as well as inefficient and the convergence of the final generative model strongly depends on the previous generators; 2) The quality of precursor images is ignored by this architecture; 3) Multiple discriminator networks are required to be trained. To this end, we propose an efficient and effective single-stage framework called Fusion-S2iGan to yield perceptually plausible and semantically consistent image samples on the basis of given spoken descriptions. Fusion-S2iGan introduces a visual+speech fusion module (VSFM), constructed with a pixel-attention module (PAM), a speech-modulation module (SMM) and a weighted-fusion module (WFM), to inject the speech embedding from a speech encoder into the generator while improving the quality of synthesized pictures. Fusion-S2iGan spreads the bimodal information over all layers of the generator network to reinforce the visual feature maps at various hierarchical levels in the architecture. We conduct a series of experiments on four benchmark data sets, i.e., CUB birds, Oxford-102, Flickr8k and Places-subset. The experimental results demonstrate the superiority of the presented Fusion-S2iGan compared to the state-of-the-art models with a multi-stage architecture and a performance level that is close to traditional text-to-image approaches.

The Procurement and Supply of hospital isolation gowns (IGs) pose significant challenges, including the potential for sudden increases in demand, the necessity of maintaining high-quality gowns, and the complexity of the supply process. One potential solution to these challenges is the investment in reusable IGs, which may seem financially infeasible due to their initial purchasing price. However, it can provide long-term financial and environmental benefits. In this research, a Simulation Optimization (SO) framework is utilized to model and analyze various product portfolio selection strategies, considering both financial and environmental perspectives, and to determine the optimal strategy for meeting both financial and environmental objectives. The proposed strategy is implemented to the problem based on obtained Life Cycle Assessment and market data.

a. The original study: Due to divergent results in replications of well-known studies in the social and biomedical sciences and perceived problems in research practices (Maxwell, Lau, and Howard 2015), attention to the importance of replication studies is growing. Replication studies in the humanities is a new field (e.g. Aguinas & Angelo 2019, Holbrook et al. 2019, Peels 2019, Dupre et al. eds. 2020). Testing how best to define, conceptualize, and implement replication studies in the humanities, both in terms of direct as well as conceptual replications, is currently being explored. This project investigates how a replication of a cornerstone historical study on the interaction of science and religion can be achieved. Towards this end a chapter of John Hedley Brooke’s 1991 book Science and Religion: some historical perspectives, entitled “The Parallel between Scientific and Religious Reform” was chosen by an interdisciplinary group of scholars at the VU Amsterdam who are interested in exploring (carrying out and reflecting on) replication studies in the humanities. The rationales for the choice of this chapter for this project were threefold: 1. Brooke’s 1991 book has been central in the study of how science and religion interact; 2. Brooke supplemented this work with a methodological appendix that facilitates replication; 3. the thesis Brooke presents in this chapter is a testable hypothesis which is suitable to both direct and conceptual replications. Brooke’s 1991 book can be characterized as catalyzing the “complexity thesis”—a perspective that contends that overarching unidimensional paradigms of conflict, separation or harmony are not born out by a detailed historical analyses (Numbers 1992). As Brooke explains in the book’s introductory chapter: Serious scholarship in the history of science has revealed so extraordinarily rich and complex a relationship between science and religion in the past that general theses are difficult to sustain. The real lesson turns out to be complexity…Conflicts allegedly about science and religion may turn out to be between rival scientific interests, or conversely between rival theological factions. Issues of political power, social prestige and intellectual authority have repeatedly been at stake…The purpose of this book…is to display the diversity, the subtlety, and ingenuity of the methods employed, both by apologists for science and for religion, as they have wrestled with the fundamental questions concerning their relationship with nature and with God. (p. 5) In chapter three, Brooke raises questions particularly regarding the connections that have been drawn between the scientific revolution and the Protestant reformation. The basic idea is that a mind-set that was open to reform and renewal in the field of religion (as the Protestant mind-set in the sixteenth and seventeenth century) may also have been open to reform and renewal in the study of nature, thus prima facie being sympathetic towards the newly proposed scientific method(s). Brooke uses responses to the heliocentric theory defended by Copernicus to test this hypothesis, comparing receptivity to heliocentrism among Protestants and Catholics; and thereby to illustrate that observing a direct connection between reform in religion and in science is more complicated than many had previously suggested. “The results of the test are instructive, but as much for the complications they reveal as for any neat conclusion” (p. 83). The findings that Brooke then goes on to present are threefold: a) there are too many complications to allow the conclusion that individuals were more or less likely to be receptive to Copernican thought depending on whether they were Protestant or Catholic; b) there are circumstances that could have made it appear as if Protestants were more supportive of the new science, for instance, if individuals were sympathetic they were more likely to enjoy freedom in publicizing their views if they were Protestant; and c) “the fortunes of the new cosmology were more deeply affected by the antagonism between Catholic and Reformed Christianity than by any peculiarities of either. In this sense there was an indirect effect of religion on science” (p.89). b. The replication: The current pre-registration proposes testing whether his conclusions hold in a direct replication. A direct replication follows the same research methods as the original study and takes into account the original sources, which turns it into a direct replication and not a reproduction. The present study will re-investigate Brooke’s claims by incorporating newly made available data as well (see below). In parallel to the study described here, a distinct conceptual replication will also be carried out by our team in which Jewish responses to heliocentrism will be examined (see https://osf.io/64gu3 ). It is hoped that by conducting both a direct and conceptual replication in parallel, new insights can be gleaned, both regarding the content of the studies, as well as regarding the process of replication studies in history. Regarding this latter topic some of the questions include: what is the difference between a conceptual replication and comparative historical studies that occur organically within the discipline? Are there advantages to engaging in both types of studies, and if so what would these be? Hence, while the bulk of the emphasis in this project will be on the replication itself.

The T5 model and its unified text-to-text paradigm contributed in advancing the state-of-the-art for many natural language processing tasks. While some multilingual variants of the T5 model have recently been introduced, their performances were found to provide suboptimal performances for languages other than English if compared to monolingual variants. We are motivated by these findings to introduce IT5, the first family of encoder-decoder transformer models pretrained specifically on Italian. We perform a thorough cleaning of a web-crawled Italian corpus including more than 40 billion words and use it to pretrain three IT5 models of different sizes. The performance of IT5 models and their multilingual counterparts is then evaluated on a broad range of natural language understanding and generation benchmarks for Italian. We find the monolingual IT5 models to provide the best scale-to-performance ratio across tested models, consistently outperforming their multilingual counterparts and setting a new state-of-the-art for most Italian conditional language generation tasks. Comment: 13 pages, 7 tables, 1 figure. Code and checkpoints available: https://github.com/gsarti/it5

Behavioural public policy is increasingly interested in scaling–up experimental insights to deliver systemic changes. Recent evidence shows some forms of individual behaviour change, such as nudging, are limited in scale. We argue, we can scale up individual behaviour change by accounting for nuanced social complexities in which human responses to behavioural public policies are situated. We introduce the idea of the “social brain”, as a construct to help practitioners and policymakers facilitate a greater social transmission of pro–social behaviours. The social brain encompasses complex human relationships interacted with elements of the choice architecture. It includes three main components: (1) individual actors, representing nodes in the social brain, who interact with other actors through (2) verbal and non-verbal cues, and who are affected by the (3) physical environment in which they belong. Ignoring the social brain runs the risk of fostering localised behavioural changes, through individual actors, which are neither scalable nor lasting. We identify pathways to facilitate changes in the social brain: either through path dependencies or critical mass shifts in individual behaviours, moderated by the brain’s property of social cohesion and multiplicity of situational and dispositional factors. In this way, behavioural changes stimulated in one part of the social brain can reach other parts and evolve dynamically. We recommend designing public policies that engage different parts of the social brain as a whole.