The emergence of attention-based architectures has led to significant improvements in the performance neural sequence-to-sequence models for text summarization. Although these have proved be effective summarizing English-written documents, their portability other languages is limited thus leaving plenty room improvement. In this paper, we present BART-IT, a model, based on BART architecture that specifically tailored Italian language. model pre-trained large corpus Italian-written pieces...

Formal communications such as public calls, announcements, or regulations are supposed to exhibit respect for diversity in terms of gender, race, age, and disability. However, human writers often lack adequate inclusive writing skills. For instance, they tend overuse the masculine a neutral form, mainly because self-trained on biased text examples. To overcome this issue, we propose leverage Generative Artificial Intelligence support language writing. Focusing formal Italian communications,...

This work is concerned with devising a robust Parkinson's (PD) disease detector from speech in real-world operating conditions using (i) foundational models, and (ii) enhancement (SE) methods. To this end, we first fine-tune several foundational-based models on the standard PC-GITA (s-PC-GITA) clean data. Our results demonstrate superior performance to previously proposed models. Second, assess generalization capability of PD extended (e-PC-GITA) recordings, collected operative conditions,...

Recognizing human emotions from videos requires a deep understanding of the underlying multimodal sources, including images, audio, and text. Since input data sources are highly variable across different modality combinations, leveraging multiple modalities often ad hoc fusion networks. To predict emotional arousal person reacting to given video clip we present ViPER, architecture modality-agnostic transformer based model combine frames, audio recordings, textual annotations. Specifically,...

Text Style Transfer (TST) is a relevant branch of natural language processing that aims to control the style attributes piece text while preserving its original content. To address TST in absence parallel data, Cycle-consistent Generative Adversarial Networks (CycleGANs) have recently emerged as promising solutions. Existing CycleGAN-based approaches suffer from following limitations: (1) They apply self-supervision, based on cycle-consistency principle, latent space. This approach turns out...

A large portion of user-generated content published on the Web consists opinions and reviews products, services, places in textual form. Many travellers tourists routinely rely such to drive their choices, shaping trips visits any place earth, specifically select hotels cities. In context hospitality management, a challenging research problem is identify effective strategies explain hotel ratings correlation with urban context. Under this umbrella, paper investigates use sentence-based...

Timeline summarization aims at presenting long news stories in a compact manner. State-of-the-art approaches first select the most relevant dates from original event timeline then produce per-date summaries. Date selection is driven by either content or date-level references. When coping with complex data, characterized inherent flow redundancy, this pipeline may encounter issues both date and due to limited use of no high-level temporal references (e.g., past month). This paper proposes...

The use of distributed vector representations words in Natural Language Processing has become established.To tailor general-purpose spaces to the context under analysis, several domain adaptation techniques have been proposed.They all require sufficiently large document corpora tailored target domains.However, cross-lingual NLP domains both enough domainspecific and pre-trained domain-specific word vectors are hard find for languages other than English.This paper aims at tackling aforesaid...

Preserving diversity and inclusion is becoming a compelling need in both industry academia. The ability to use appropriate forms of writing, speaking, gestures not widespread even formal communications such as public calls, announcements, official reports, legal documents. improper linguistic expressions can foment unacceptable exclusion, stereotypes well verbal violence against minorities, including women. Furthermore, existing machine translation tools are designed generate inclusive...

Podcasts are becoming an increasingly popular way to share streaming audio content. Podcast summarization aims at improving the accessibility of podcast content by automatically generating a concise summary consisting text/audio extracts. Existing approaches either extract short snippets means speech techniques or produce abstractive summaries transcription disregarding audio. To leverage multimodal information hidden in episodes we propose end-to-end architecture for extractive that encodes...

Cet article présente le projet E-MIMIC, une application qui vise à éliminer les préjugés et la non-inclusion dans textes administratifs rédigés pays européens, commencer par ceux sont langues romanes. Il méthodologie conçue partir de critères discursifs inspirés l’analyse du discours française utilisés pour étiqueter un corpus documents institutionnels, l’apprentissage profond des réseaux neuronaux. Des architectures modélisation profonde langage exploitées identifier automatiquement...

Abstract Scientific articles often include in-text citations quoting from external sources. When the cited source is an article, citation context can be analyzed by exploring article full-text. To quickly access key information, researchers are interested in identifying sections of that most pertinent to text surrounding citing article. This paper first performs a data-driven analysis correlation between textual content and snippet where placed. The results show title abstract likely highly...