1,721,223 research outputs found
BaBIEs: A Benchmark for the Linguistic Evaluation of Italian Baby Language Models
The possibility of comparing the linguistic competence of Language Models (LMs) to that of children has gained growing attention lately, raising the need for effective tools for evaluating both the former and the latter. To this purpose, we developed a resource for the linguistic evaluation of BabyLMs, which are LMs trained on datasets that comparable to the linguistic stimulus received by children. This resource adapts four standardized tests for the evaluation of linguistic skills of Italianspeaking children (BVL, TROG-2, TCGB-2 and Peabody). To verify the effectiveness of our benchmark, we administered it to
Minerva, a LLM pretrained from scratch on Italian. Our results indicate that Minerva struggles to master certain linguistic aspects, achieving an age-equivalent score of 4 years, and that the type of task administered affects the model’s performance
Verbs of motion and the argument-adjunct continuum in Italian
• Discussion of the argument-adjunct status of the locative phrase optionally co-occurring with verbs of motion in Italian, in relation to the interplay of syntactic criteria (e.g., obligatoriness, latency)/Definite Null Instantiation), different order constraints) with semantic parameters: (i) the degree of lexicalization of the direction of movement in the verbal roots, (ii) the event structure template of predicates, (iii) the inherent and relational characteristics of the subject, (iv) the semantics of the preposition(s
Gradience in subcategorization? Locative phrases with Italian verbs of motion
• Discussion of the relationship between a (verbal) head and its dependent(s), with reference to the argument-adjunct status of the locative phrase optionally co-occurring with verbs of motion in Italian, both bounded, i.e., directed motion – e.g., andare 'to go', arrivare 'to arrive', venire 'to come' – and unbounded, i.e., manner of motion – e.g., nuotare 'to swim', galleggiare 'to float', rotolare 'to roll', correre 'to run'
Complex Fluids in Fractured Geological Media for Enhanced Heat Transfer - GEONEAT
Engineered shear-thinning (ST) fluids exhibit diverse applications across various fields. The MSCA-funded GEONEAT project is dedicated to investigating the potential of these engineered fluids in enhancing geothermal systems and heat tracer tests. It merges methodologies from applied mathematics and hydrology to advance heat recovery efficiency and deduce structural parameters within deep hot reservoirs. The project outlines two scientific objectives: firstly, to assess the impact of engineered fluids on coupled flow and heat transfers in fractured media; secondly, to extract structural information regarding fractured media through inverse modelling and field measurements of heat exchange. The goal involves the development of a multi-scale coupled flow and heat transport model within fractured geological formations
Story of a construction: Statistical and distributional analysis of the development of the Italian gerundival construction
This study investigates the diachronic development of the Italian gerundival periphrases stare/andare/venire + gerund using a corpus-based approach. The productivity and semantic structure of these constructions are explored over eight centuries, based on data extracted from the Google Ngram Corpus. Combining insights from Construction Grammar and Distributional Semantics, the study tracks changes in verb-collocation patterns and constructs distributional semantic spaces to model shifts in meaning and usage. The findings reveal that constructional productivity is shaped by prior usage and analogical patterns, with each construction exhibiting different degrees of schematicity
“The Talmud System: a Collaborative Web Application for the Translation of the Babylonian Talmud into Italian.”
The article introduces the Talmud System, a collaborative web application for the translation of the Babylonian
Talmud into Italian. The system, which was developed in the context of the “Progetto Traduzione del Talmud Babilonese,” was designed to improve the experience of collaborative translation using Computer-Assisted Translation technologies and providing a rich environment for the creation of comments and the annotation of text on a linguistic and semantic basis
Drainage of Gravity Currents from an Edge and a Permeable Substrate in a Porous Medium or Fracture with Variable Properties
Gravity-driven flow in porous media has been extensively investigated in recent years in connection with numerous environmental and industrial applications, including seawater intrusion, oil recovery, penetration of drilling fluids into reservoirs, contaminant migration such as NAPL spreading in shallow aquifers, and carbon dioxide sequestration in deep subsurface formations. Analytical and numerical solutions are available for various geometries and boundary conditions, including different drainage mechanisms. Their use can be extended to gravity-driven flow in narrow vertical fractures or cracks via the well-known Hele-Shaw (HS) analogy between parallel plate and porous media flow, with the aperture b squared being the analog of permeability k according to k = b2/12. The propagation of gravity currents in porous and fractured media is mainly governed by the interplay between viscous and buoyancy forces, typically with negligible inertial effects, and is affected by spatial heterogeneity of medium properties; permeability, porosity, and aperture gradients have been shown to affect the propagation distance and shape of gravity currents, with practical implications for remediation and storage. This paper is interested in the dynamics of gravity currents propagating in a porous medium with spatial properties varying parallel to the flow direction under the coupled drainage mechanisms of a fixed edge and a permeable substrate. Simultaneous permeability and porosity gradients parallel to the flow are considered: this is equivalent to a fracture with a horizontally variable aperture, as the Hele-Shaw analogy necessarily accounts for both permeability and porosity gradients. We consider a Newtonian fluid with a density of ρ+Δρ intruding into a porous medium and advancing in a fluid of density ρ under the sharp interface approximation, where a no-flow boundary condition is considered at the origin x = 0. We assume that the fluid drains away simultaneously from the permeable base and an edge and neglect vertical velocities in a long and thin current; this implies vertical equilibrium and, in turn, a hydrostatic pressure distribution within the advancing current. The final assumption is of vanishing height of the current at the draining edge after a relatively short adjustment time, favoured by increased permeability/porosity or aperture along the flow direction. Under these assumptions, a semi-analytical solution is derived for the height of the current h(x, t) in a self-similar form, valid as a late-time approximation modelling the drainage phenomenon after the influence of the initial condition has vanished. This allows transforming the nonlinear PDE governing the flow into a nonlinear ODE amenable to a numerical solution. Besides, the current profile yields the amount of fluid loss through each of the drainage mechanisms. Results are discussed as a function of model parameters, and an analysis of the conditions required to avoid an unphysical or asymptotically invalid result is presented
Refining the Distributional Inclusion Hypothesis for Unsupervised Hypernym Identification
Several unsupervised methods for hypernym detection have been investigated in distributional semantics. Here we present a new approach based on a smoothed version of the distributional inclusion hypothesis. The new method is able to improve hypernym detection after testing on the BLESS dataset
MEDEA: Merging Event knowledge and Distributional vEctor Addition
The great majority of composi- tional models in distributional semantics present methods to compose distributional vectors or tensors in a representation of the sentence. Here we propose to enrich the best performing method (vector addition, which we take as a baseline) with distri- butional knowledge about events, outper- forming our baseline
Comparing Static and Contextual Distributional Semantic Models on Intrinsic Tasks: An Evaluation on Mandarin Chinese Datasets
The field of Distributional Semantics has recently undergone important changes, with the contextual representations produced by Transformers taking the place of static word embeddings models. Noticeably, previous studies comparing the two types of vectors have only focused on the English language and a limited number of models. In our study, we present a comparative evaluation of static and contextualized distributional models for Mandarin Chinese, focusing on a range of intrinsic tasks. Our results reveal that static models remain stronger for some of the classical tasks that consider word meaning independent of context, while contextualized models excel in identifying semantic relations between word pairs and in the categorization of words into abstract semantic classes
- …
