63,896 research outputs found
On Automatic Prominence Detection for German
No full textTamburini F, Wagner P. On Automatic Prominence Detection for German. In: Proceedings of Interspeech 2007. 2007: 1809-1812.Perceptual prominence is an important indicator of a word's and syllable's lexical, syntactic,
semantic and pragmatic status in a discourse. Its automatic annotation would be a valuable
enrichment of large databases used in unit selection speech synthesis and speech recognition.
While much research has been carried out on the interaction between prominence and
acoustic factors, little progress has been made in its automatic annotation. Previous
approaches to German relied on linguistic features in prominence detection, but a purely
acoustic method would be advantageous. We applied an algorithm to German data that had
been previously used for English and Italian. Both the algorithm and the data annotation
encode prominence as a continuous rather than a categorical parameter. First results are
encouraging, but again show that prominence perception relies on linguistic expectancies as
well as acoustic patterns. Also, our results further strengthen the view that force accents are a
more reliable cue to prominence than pitch accents in German
Objective, Subjective and Linguistic Roads to Perceptual Prominence. How are they compared and why?
No full textWagner P, Tamburini F, Windmann A. Objective, Subjective and Linguistic Roads to Perceptual Prominence. How are they compared and why? In: Proceedings of Interspeech 2012. 2012: 2386-2389
Curated Data does not mean Representative Data when training Large Language Models: an Experiment using Representative Data for Italian
No full textIt is widely accepted in literature that data curation is the first step for a successful pretraining of Large, and Small, Language Models (LLMs). Datasets generally fall into two categories: open datasets are publicly available, fostering transparency, reproducibility, and community-driven improvement, but they often face limitations in scale, diversity, and quality. Closed datasets, typically curated by private entities, can offer greater scale, higher quality, and proprietary data sources, yet they raise concerns around transparency, bias auditing, and public accountability.
This paper presents an experiment aimed at quantitatively measuring the improvements provided by representative datasets for LLM pretraining. We pretrained two small LLMs under the same experimental conditions as the corresponding Italian reference models from the Minerva family, evaluated their performance on standard benchmarks, and used LLM-as-a- Judge to assess the Fluency, Coherence, and Relevance of generated texts on specific tasks. The results support the idea that, while open science and open datasets are important goals, representative corpora, even if closed, are more suitable for LLM pretraining, as they enable better performance under identical experimental conditions
Nilpotent groups of semilinear transformations which are monomial
No full textLet H be a nilpotent subgroup of ΓL_n (q) = GL_n (q), where φ denotes the field automorfism induced by the Frobenius map. We give a condition on the primes dividing |H ∩ GL_n (q)| under which H is conjugate to a subgroup of the generalized monomial group Diag_n (F∗_q ) Sym(n).We show an application of this result to the determination of Carter subgroups of finite groups. </p
Automatic Prominence Annotation of a German Speech Synthesis Corpus: Towards Prominence-Based Prosody Generation for Unit Selection Synthesis
No full textWindmann A, Wagner P, Tamburini F, Arnold D, Oertel C. Automatic Prominence Annotation of a German Speech Synthesis Corpus: Towards Prominence-Based Prosody Generation for Unit Selection Synthesis. In: Proceedings of the 7th ISCA Tutorial and Research Workshop on Speech Synthesis. 2010: 377-382
A property of the normalizers of Sylow p-subgroups in simple groups
No full textQuaderni del Seminario Matematico di Bresci
Comparison of fibers in geopolymer matrix for structural reinforcement of masonry (FRGP): compatibility, reactivity, durability
No full textGeopolymer matrix bonded fibre nets of UHTSS-steel, carbon, basalt and AR-glass were
prepared on soft mud bricks. Brick reinforcement was measured by three-point flexural tests and fibrematrix-
substrate adhesion by pull-off tests. The durability of the reinforcements was tested by aging and
freeze-thaw cycles. Alkali resistance of fibres was tested in different alkaline solution and comprised
weight loss and tensile strength
Groundwater temperature as a natural tracer to characterize hydraulic behaviour and geometry of carbonate aquifers:Mt. Nerone karst system, central Italy
No full textTemperature variations in groundwater discharge from the Giordano karst system (northern Apennines) have been observed continuously for six months (from January 2016 to July 2016), to determine the flow patterns and geometrical properties of aquifer. The karst system discharge about 32.2 L/s on average during over the hydrological year and it’s characterized by two outlets: a basal continuous spring (BCS) with a mean discharge of 8.9 L/s and an upper-intermittent spring (UIS) with a mean discharge of 23.3 L/s, located about 60 m above the previous one. Results show significant difference in the two spring outlets between the time lags as well as non-simultaneous and not analogue responses of temperature (T) to the same recharge events: temperature of UIS range from 9.7 to 10.7°C whereas temperature of BCS remain rather stable, ranging from 9.8 to 9.9°C. This data suggests
a stratification of the water along the aquifer probably associated by different residence times and linked to the structural organization of karst aquifer (conduit and fracture networks): deeper and oldest water in the BCS and youngest water in the UIS
- …
