Learning Methods and Algorithms for Semantic Text Classification across Multiple Domains

Information is nowadays a key resource: machine learning and data mining techniques have been developed to extract high-level information from great amounts of data. As most data comes in form of unstructured text in natural languages, research on text mining is currently very active and dealing with practical problems. Among these, text categorization deals with the automatic organization of large quantities of documents in priorly defined taxonomies of topic categories, possibly arranged in large hierarchies. In commonly proposed machine learning approaches, classifiers are automatically trained from pre-labeled documents: they can perform very accurate classification, but often require a consistent training set and notable computational effort. Methods for cross-domain text categorization have been proposed, allowing to leverage a set of labeled documents of one domain to classify those of another one. Most methods use advanced statistical techniques, usually involving tuning of parameters. A first contribution presented here is a method based on nearest centroid classification, where profiles of categories are generated from the known domain and then iteratively adapted to the unknown one. Despite being conceptually simple and having easily tuned parameters, this method achieves state-of-the-art accuracy in most benchmark datasets with fast running times. A second, deeper contribution involves the design of a domain-independent model to distinguish the degree and type of relatedness between arbitrary documents and topics, inferred from the different types of semantic relationships between respective representative words, identified by specific search algorithms. The application of this model is tested on both flat and hierarchical text categorization, where it potentially allows the efficient addition of new categories during classification. Results show that classification accuracy still requires improvements, but models generated from one domain are shown to be effectively able to be reused in a different one.

La ricerca traslazionale in farmacologia gastroenterologica

Il presente studio si concentra sull’analisi degli aspetti traslazionali nella ricerca farmacologica applicata alla Gastroenterologia. La trattazione si articola in due parti: una prima elaborazione teorica, che permette di inquadrare nel contesto della ricerca traslazionale il razionale scientifico ed etico alla base delle attività sperimentali eseguite durante il triennio; una seconda parte, nella quale si riportano i metodi, i risultati e le osservazioni conclusive derivanti dallo studio sperimentale. Nella prima parte vengono analizzate alcune caratteristiche delle procedure, adottate nella ricerca in ambito farmacologico gastrointestinale, che permettono di ottenere un dato verosimile derivabile da modelli diversi rispetto all’organismo umano. Sono inclusi nella trattazione gli aspetti etici dell’utilizzo di alcuni modelli animali di patologie intestinali organiche e funzionali in relazione al loro grado di predittività rispetto alla realtà sperimentale clinica. Nella seconda parte della trattazione, viene presentato uno studio esplorativo tissutale multicentrico sul ruolo del sistema oppioide e cannabinoide nella sindrome dell’intestino irritabile (IBS). Obiettivo dello studio è la valutazione dell’espressione e la localizzazione del recettore oppioide µ (µOR), del suo ligando β endorfina (β-END) e del recettore cannabinoide 2 (CB2) nei pazienti con IBS ad alvo costipato (IBS-C) e diarroico (IBS-D), ed in soggetti sani (HC). I dati ottenuti indicano un’implicazione del sistema oppioide e cannabinoide nella risposta immune alterata riscontrata nei pazienti con IBS ed in particolare nel sottogruppo IBS-C. La presente trattazione suggerisce come la creazione di nuovi sistemi di indagine sempre più validi da un punto di vista traslazionale possa dipendere, almeno in parte, dalla capacità di integrare realtà disciplinari, tecnologie ed esperienze metodologiche diverse nel contesto della ricerca in campo biomedico e farmacologico ed in particolare tramite un mutuo scambio di informazioni tra realtà clinica e ricerca di base

Meccanizzazione dei vigneti allevati a doppia cortina e a cordone libero

La ricerca svolta ha voluto approfondire le possibilità offerte dai sistemi di allevamento dei vigneti a Doppia Cortina (GDC) e a Cordone Libero nei riguardi della meccanizzazione. La ricerca ha considerato gli interventi di potatura invernale, di gestione della chioma (spollonatura, cimatura, defogliazione e pettinatura della doppia cortina) e di vendemmia. Un’operazione particolarmente seguita è stata la potatura invernale realizzando differenti livelli di meccanizzazione. Tutti gli interventi sono stati eseguiti sia manualmente che meccanicamente, confrontando i tempi d’impiego, la qualità del lavoro svolto e gli impegni di manodopera. I risultati sono stati sintetizzati in una valutazione economica, ipotizzando differenti livelli di costo della manodopera impiegata, per ottenere giudizi di convenienza per i singoli interventi e per costruire una valutazione completa e più organica della linea di lavoro proposta. Nelle due forme d’allevamento la meccanizzazione della potatura invernale e della gestione della chioma hanno rispettato pienamente gli obbiettivi tecnici prefissati, dimostrando di essere un valido mezzo per ridurre tempi e costi di gestione. Per questi interventi l’acquisto delle macchine risulta conveniente anche per vigneti di piccola dimensione. Ancor più evidenti in queste due forme d’allevamento sono i vantaggi economici offerti dalla vendemmia meccanica, realizzata con pochi maltrattamenti e perdite di prodotto. La tendenza a meccanizzare integralmente gli interventi di gestione del ciclo colturale della vite, può essere nei prossimi anni un motivo di interesse e di scelta nella realizzazione di nuovi impianti con queste due forme di allevamento, che hanno dimostrato di essere un’espressione completa di sinergia tra macchina e pianta.

Modellazione e simulazione di sistemi meccanici ad elevato numero di gradi di libertà mediante metodi non convenzionali e sistemi CAD

Recent developments in piston engine technology have increased performance in a very significant way. Diesel turbocharged/turbo compound engines, fuelled by jet fuels, have great performances. The focal point of this thesis is the transformation of the FIAT 1900 jtd diesel common rail engine for the installation on general aviation aircrafts like the CESSNA 172. All considerations about the diesel engine are supported by the studies that have taken place in the laboratories of the II Faculty of Engineering in Forlì. This work, mostly experimental, concerns the transformation of the automotive FIAT 1900 jtd – 4 cylinders – turbocharged – diesel common rail into an aircraft engine. The design philosophy of the aluminium alloy basement of the spark ignition engine have been transferred to the diesel version while the pistons and the head of the FIAT 1900 jtd are kept in the aircraft engine. Different solutions have been examined in this work. A first V 90° cylinders version that can develop up to 300 CV and whose weight is 30 kg, without auxiliaries and turbocharging group. The second version is a development of e original version of the diesel 1900 cc engine with an optimized crankshaft, that employ a special steel, 300M, and that is verified for the aircraft requirements. Another version with an augmented stroke and with a total displacement of 2500 cc has been examined; the result is a 30% engine heavier. The last version proposed is a 1600 cc diesel engine that work at 5000 rpm, with a reduced stroke and capable of more than 200 CV; it was inspired to the Yamaha R1 motorcycle engine. The diesel aircraft engine design keeps the bore of 82 mm, while the stroke is reduced to 64.6 mm, so the engine size is reduced along with weight. The basement weight, in GD AlSi 9 MgMn alloy, is 8,5 kg. Crankshaft, rods and accessories have been redesigned to comply to aircraft standards. The result is that the overall size is increased of only the 8% when referred to the Yamaha engine spark ignition version, while the basement weight increases of 53 %, even if the bore of the diesel version is 11% lager. The original FIAT 1900 jtd piston has been slightly modified with the combustion chamber reworked to the compression ratio of 15:1. The material adopted for the piston is the aluminium alloy A390.0-T5 commonly used in the automotive field. The piston weight is 0,5 kg for the diesel engine. The crankshaft is verified to torsional vibrations according to the Lloyd register of shipping requirements. The 300M special steel crankshaft total weight is of 14,5 kg. The result reached is a very small and light engine that may be certified for general aviation: the engine weight, without the supercharger, air inlet assembly, auxiliary generators and high pressure body, is 44,7 kg and the total engine weight, with enlightened HP pump body and the titanium alloy turbocharger is less than 100 kg, the total displacement is 1365 cm3 and the estimated output power is 220 CV. The direct conversion of automotive piston engine to aircrafts pays too huge weight penalties. In fact the main aircraft requirement is to optimize the power to weight ratio in order to obtain compact and fast engines for aeronautical use: this 1600 common rail diesel engine version demonstrates that these results can be reached.

Indagine sull’influenza di differenti sistemi d’allevamento sui connotati nutrizionali, tecnologici, sensoriali e di freschezza dell’orata (Sparus aurata) allevata in Italia

On account of the commercial importance of gilthead sea bream (Sparus aurata) in Italy the aim of the present study is the evaluation of the quality of nutritional, technological, sensory and freshness aspects. Sea bream production is growing in the Mediterranean and the evaluation of its quality concerns both producers and consumers alike. The culture system greatly influences final product quality. In Italy most of sea bream culture is carried out in cages but there is also a production in land based facilities and in lagoons. In this study external appeareance differentiations are pronounced. Different results were found for nutritional aspects, for fatty acids and for mineral content. Some differences in the freshness indices are also found. Furthermore, organoleptic differences have been described between culture system.

Studio e sviluppo di sistemi RFID in banda UHF

The recent widespread diffusion of radio-frequency identification (RFID) applications operating in the UHF band has been supported by both the request for greater interrogation ranges and greater and faster data exchange. UHF-RFID systems, exploiting a physical interaction based on Electromagnetic propagation, introduce many problems that have not been fully explored for the previous generations of RFID systems (e.g. HF). Therefore, the availability of reliable tools for modeling and evaluating the radio-communication between Reader and Tag within an RFID radio-link are needed. The first part of the thesis discuss the impact of real environment on system performance. In particular an analytical closed form formulation for the back-scattered field from the Tag antenna and the formulation for the lower bound of the BER achievable at the Reader side will be presented, considering different possible electromagnetic impairments. By means of the previous formulations, of the analysis of the RFID link operating in near filed conditions and of some electromagnetic/system-level co-simulations, an in-depth study of the dimensioning parameters and the actual performance of the systems will be discussed and analyzed, showing some relevant properties and trade-offs in transponder and reader design. Moreover a new low cost approach to extend the read range of the RFID UHF passive systems will be discussed. Within the scope to check the reliability of the analysis approaches and of innovative proposals, some reference transponder antennas have been designed and extensive measurement campaign has been carried out with satisfactory results. Finally, some commercial ad-hoc transponder for industrial application have been designed within the cooperation with Datalogic s.p.a., some guidelines and results will be briefly presented.

Meccanismi evolutivi dei parapoxvirus: caratterizzazione genomica di pseudocowpoxvirus e messa a punto di sistemi per lo studio delle ricombinazioni

The Poxviruses are a family of double stranded DNA (dsDNA) viruses that cause disease in many species, both vertebrate and invertebrate. Their genomes range in size from 135 to 365 kbp and show conservation in both organization and content. In particular, the central genomic regions of the chordopoxvirus subfamily (those capable of infecting vertebrates) contain 88 genes which are present in all the virus species characterised to date and which mostly occur in the same order and orientation. In contrast, however, the terminal regions of the genomes frequently contain genes that are species or genera-specific and that are not essential for the growth of the virus in vitro but instead often encode factors with important roles in vivo including modulation of the host immune response to infection and determination of the host range of the virus. The Parapoxviruses (PPV), of which Orf virus is the prototypic species, represent a genus within the chordopoxvirus subfamily of Poxviridae and are characterised by their ability to infect ruminants and humans. The genus currently contains four recognised species of virus, bovine papular stomatitis virus (BPSV) and pseudocowpox virus (PCPV) both of which infect cattle, orf virus (OV) that infects sheep and goats, and parapoxvirus of red deer in New Zealand (PVNZ). The ORFV genome has been fully sequenced, as has that of BPSV, and is ~138 kb in length encoding ~132 genes. The vast majority of these genes allow the virus to replicate in the cytoplasm of the infected host cell and therefore encode proteins involved in replication, transcription and metabolism of nucleic acids. These genes are well conserved between all known genera of poxviruses. There is however another class of genes, located at either end of the linear dsDNA genome, that encode proteins which are non-essential for replication and generally dictate host range and virulence of the virus. The non-essential genes are often the most variable within and between species of virus and therefore are potentially useful for diagnostic purposes. Given their role in subverting the host-immune response to infection they are also targets for novel therapeutics. The function of only a relatively small number of these proteins has been elucidated and there are several genes whose function still remains obscure principally because there is little similarity between them and proteins of known function in current sequence databases. It is thought that by selectively removing some of the virulence genes, or at least neutralising the proteins in some way, current vaccines could be improved. The evolution of poxviruses has been proposed to be an adaptive process involving frequent events of gene gain and loss, such that the virus co-evolves with its specific host. Gene capture or horizontal gene transfer from the host to the virus is considered an important source of new viral genes including those likely to be involved in host range and those enabling the virus to interfere with the host immune response to infection. Given the low rate of nucleotide substitution, recombination can be seen as an essential evolutionary driving force although it is likely underestimated. Recombination in poxviruses is intimately linked to DNA replication with both viral and cellular proteins participate in this recombination-dependent replication. It has been shown, in other poxvirus genera, that recombination between isolates and perhaps even between species does occur, thereby providing another mechanism for the acquisition of new genes and for the rapid evolution of viruses. Such events may result in viruses that have a selective advantage over others, for example in re-infections (a characteristic of the PPV), or in viruses that are able to jump the species barrier and infect new hosts. Sequence data related to viral strains isolated from goats suggest that possible recombination events may have occurred between OV and PCPV (Ueda et al. 2003). The recombination events are frequent during poxvirus replication and comparative genomic analysis of several poxvirus species has revealed that recombinations occur frequently on the right terminal region. Intraspecific recombination can occur between strains of the same PPV species, but also interspecific recombination can happen depending on enough sequence similarity to enable recombination between distinct PPV species. The most important pre-requisite for a successful recombination is the coinfection of the individual host by different virus strains or species. Consequently, the following factors affecting the distribution of different viruses to shared target cells need to be considered: dose of inoculated virus, time interval between inoculation of the first and the second virus, distance between the marker mutations, genetic homology. At present there are no available data on the replication dynamics of PPV in permissive and non permissive hosts and reguarding co-infetions there are no information on the interference mechanisms occurring during the simultaneous replication of viruses of different species. This work has been carried out to set up permissive substrates allowing the replication of different PPV species, in particular keratinocytes monolayers and organotypic skin cultures. Furthermore a method to isolate and expand ovine skin stem cells was has been set up to indeep further aspects of viral cellular tropism during natural infection. The study produced important data to elucidate the replication dynamics of OV and PCPV virus in vitro as well as the mechanisms of interference that can arise during co-infection with different viral species. Moreover, the analysis carried on the genomic right terminal region of PCPV 1303/05 contributed to a better knowledge of the viral genes involved in host interaction and pathogenesis as well as to locate recombination breakpoints and genetic homologies between PPV species. Taken together these data filled several crucial gaps for the study of interspecific recombinations of PPVs which are thought to be important for a better understanding of the viral evolution and to improve the biosafety of antiviral therapy and PPV-based vectors.