1731 research outputs found
Sort by
Exploiting the Volatile Nature of Data and Information in Evolving Repositories and Systems with User Generated Content
No full textModern technological advances have created a plethora of an extremely large, highly heterogeneous and distributed collection of datasets that are highly volatile. This volatile nature makes their understanding, integration and management a challenging task.
One of the first challenging issues is to create the right models that will capture not only the changes that have taken place on the values of the data but also the semantic evolution of the concepts that the data structures represent. Once this information has been captured, the right mechanisms should be put in place to enable the exploitation of the evolution information in query formulation, reasoning, answering and representation. Additionally, the continuously evolving nature of the data hinders the ability of determining the quality of the data that is observed at a specific moment, since there is a great deal of uncertainty on whether this information will remain as is. Finally, an important task in this context, known as information filtering, is to match a specific piece of information which is recently updated (or added) in a repository to a user or query at hand.
In this dissertation, we propose a novel framework to model and query data which have the explicit evolution relationships among concepts. As a query language we present an expressive evolution graph traversal query language which is tested on a number of real case scenarios: the history of Biotechnology, the corporate history of US companies and others. In turn, to support query evaluation we introduce an algorithm using the idea of finding Steiner trees on graphs which is capable of computing answers on-the-fly taking into account the evolution connections among concepts.
To address the problem of data quality in user generated repositories (e.g. Wikipedia) we present a novel algorithm which detects individual controversies by using the substitutions in the revision history of a content. The algorithm groups the disagreements between users by means of a context, i.e. the surrounding content, and by applying custom filters. In the extensive experimental evaluation we showed that the proposed ideas lead to high effectiveness on a various sources of controversies.
Finally, we exploit the problem of producing recommendations in evolving repositories by focusing on the cold start problem, i.e. when no or little past information about the users and/or items is given. In the dissertation we present a number of novel algorithms which cope with the cold-start by leveraging the item features using the k-neighbor classifier, Naive Bayes classifier and maximum entropy principle. The obtained results enable recommender systems to operate in rapidly updated domains such that news, university courses and social data
Development of enhanced double-sided 3D radiation sensors for pixel detector upgrades at HL-LHC
No full textThe upgrades of High Energy Physics (HEP) experiments at the Large Hadron Collider (LHC) will call for new radiation hard technologies to be applied in the next generations of tracking devices that will be required to withstand extremely high radiation doses. In this sense, one of the most promising approaches to silicon detectors, is the so called 3D technology. This technology realizes columnar electrodes penetrating vertically into the silicon bulk thus decoupling the active volume from the inter-electrode distance. 3D detectors were first proposed by S. Parker and collaborators in the mid ’90s as a new sensor geometry intended to mitigate the effects of radiation damage in silicon. 3D sensors are currently attracting growing interest in the field of High Energy Physics, despite their more complex and expensive fabrication, because of the much lower operating voltages and enhanced radiation hardness. 3D technology was also investigated in other laboratories, with the intent of reducing the fabrication complexity and aiming at medium volume sensor production in view of the first upgrades of the LHC experiments. This work will describe all the efforts in design, fabrication and characterization of 3D detectors produced at FBK for the ATLAS Insertable B-Layer, in the framework of the ATLAS 3D sensor collaboration. In addition, the design and preliminary characterization of a new batch of 3D sensor will also be described together with new applications of 3D technology
Visual Saliency Detection and its Application to Image Retrieval
No full textPeople perceive any kind of information with different level of attention and involvement. It is due to the way how our brain functions, redundancy and importance of the perceived data. This work deals with visual information, in particular with images.
Image analysis and processing is often requires running computationally expensive algorithms. The knowledge of which part of an image is important over other parts allows for reduction of data to be processed. Besides computational cost a broad variety of applications, including image compression, quality assessment, adaptive content display and rendering, can benefit from this kind of information. The development of an accurate visual importance estimation method may bring a useful tool for image processing domain and that is the main goal for this work. In the following two novel approaches to saliency detection are presented. In comparison to previous works in this field the proposed approaches tackles saliency estimation on the object-wise level. In addition, one of the proposed approach solves saliency detection problem through modelling 3-D spatial relationships between objects in a scene.
Moreover, a novel idea of the application of saliency to diversification of image retrieval results is presented
Integrated Methodologies Based on Structural Health Monitoring for the Protection of Cutural Heritage Buildings
No full textIn the last decades the need for an effective seismic protection and vulnerability reduction of strategic structures and particularly the architectural heritage determined a growing interest in Structural Health Monitoring (SHM) as a measure of passive mitigation of earthquake effects. The object of monitoring is to identify, locate and classify type and severity of damages induced by external actions or degradation phenomena and to assess their effects on the structural performance. In this way it is possible to take appropriate measures to reduce the danger of collapse and, when necessary, perform strengthening interventions to improve the structural and seismic capacity.
Motivated by the above reasons, this thesis aims at providing a contribution to the development of techniques and integrated methodologies, based on SHM, for the assessment and protection of Cultural Heritage (CH) buildings and monuments.
Firstly, after a detailed state of the art review on specific topics related to SHM of civil engineering structures, a new methodology for the implementation of monitoring techniques on historic masonry structures is proposed. Selected case studies, equipped with distributed sensors and acquisition systems, allowed the definition and successive validation of SHM as a knowledge-based assessment tool, implemented to evaluate intervention needs, following an incremental approach during their execution, and to control the damage states of buildings in a post-seismic scenario.
In order to maximize the benefits of SHM and optimize the entire process, dedicated software for static monitoring and automated algorithms for modal parameters identification have been developed, able to provide almost real time information on the health state of the monitored structure.
Finally integrated procedures based on robust statistical and numerical models have been implemented to interpret and exploit SHM outputs to assess the structural conditions of the investigated CH buildings
Improving the Effectiveness of Information Extraction from Biomedical Text
No full textInformation extraction (IE) is the task which aims at automatically extracting specific target information from texts by means of various natural language processing (NLP) and Machine Learning (ML) techniques. The huge amount of available biomedical and clinical texts is an important source of undiscovered knowledge and an interesting domain where IE techniques can be applied. Although there has been a considerable amount of work for IE on other genres of text (such as newspaper articles), results of the state-of-the-art approaches for some of the IE tasks show there is still the need of improvement. Moreover, when these IE approaches are directly applied on biomedical/clinical data, the performance drops considerably. Customization of the IE approaches with biomedical/clinical genre specific features and pre/post-processing techniques does improve the results (with respect to applying the approaches directly) but the situation is still not completely satisfactory. There are many ways to accomplish this goal (e.g. exploitation of scope of negations, discourse structure, semantic roles, etc) which are yet to be fully harnessed for the improvement of IE systems. Additional challenges come from the usage of machine learning (ML) techniques themselves. Imbalance in data distribution is quite common in many NLP (including IE) tasks. Previous studies have empirically shown that unbalanced datasets lead to poor performance for the minority class.
In this PhD research, we aim to address the open issues outlined above. We focus on three core IE tasks which are crucial for text mining: named entity recognition (NER), coreference resolution (CoRef), and relation extraction (RE).
For NER, we propose an approach for the recognition of disease entity mentions which achieves state-of-the-art performance and is later exploited as a component in our RE system. Our NER system achieves results on par with the state of the art also for other bio-entity types such as genes/proteins, species and drugs. Since the creation of manually annotated training data is a costly process, we also investigate the practical usability of automatically annotated corpora for NER and propose how to automatically improve the quality of such corpora.
CoRef, which is naturally the next step after NER, is often deemed as one of the stumbling blocs for other IE tasks such as RE. We propose a greedy and constrained CoRef approach that achieves high results in clinical texts for each individual entity mention type and for each of the four different evaluation metrics usually computed for assessing systems' performance.
As for RE, one of the fundamental characteristics of our approach is that we propose to exploit other NLP areas such as scope of negations, elementary discourse units and semantic roles. We propose a novel hybrid kernel that not only takes advantage of different types of information (syntactic, semantic, contextual, etc) but also of the different ways they can be represented (i.e. flat structure, tree, graph). Our approach yields significantly better results than the previous state-of-the-art approaches for drug-drug interaction and protein-protein interaction extraction tasks.
In each of the above tasks, we concentrate to develop pro-active IE approaches to automatically get rid of unnecessary training/test instances even before training ML models and using those models on test data. This enables better performance because of the reduction of less skewed data distribution as well as faster runtime.
We tested our NER and RE approaches on other genres of text such as newspaper articles and automatically transcribed broadcast news. The results show that our approaches are largely domain independent
Production of a nanostructured copper by Spark Plasma Sintering
No full textThe aim of the present PhD work is the study of the production of a nanostructured copper by Spark Plasma Sintering.
The nanostructured powder was produced by cryomilling an atomized powder, using a ball-to-powder ratio of 30:1 for 8h; it has a mean grain size of 19±2 nm and shows quite a high thermal stability, as shown by a DSC investigation.
The influence of temperature, particle size, pressure on the densification and sintering mechanisms as well as that of heating rate and holding time on the structural evolution has been investigated. Particle rearrangement, local deformation, bulk deformation and sintering are the SPS mechanisms occurring successively during the sintering process of the atomized copper. These mechanisms are enhanced by the peculiar heating mechanism in SPS, and the surface overheating above the melting temperature in the contact regions has been demonstrated. In the cryomilled powder, sintering occurs at much lower temperature than in the atomized powder, due to effect of the high density of structural defects on the mass transport phenomena responsible for neck growth.
The increase in heating rate tends to promote a bimodal grain size distribution (both nanomentric and ultrafine grains) while an increase in holding time increases grain size slightly. A promising combination of strength and ductility was measured on tensile specimens produced under selected conditions, and a dimpled fracture morphology was observed
Hidden Consequences of Remittances: The micro impact of remittance flows in the Albanian remittance-receiving households' behavior
No full textDuring the last two decades, South-East European countries have experienced a large increase in the number of people migrating to more developed countries. With a large portion of their population abroad, these countries are highly dependent on remittances, which in the case of Albania far exceed Foreign Direct Investments.
Using household survey data for Albania, the first part of the study compares decision-making about human capital investment in remittance-receiving households and non-remittance-receiving households. The Cox proportional hazard model is used to capture the effects of remittances.
The second part the dissertation focuses on the impact of remittances on labor market participation using propensity score matching. This part of the dissertation relies on the matching approach for the identification. The nearest neighbor and kernel estimators are used to obtain the matching results.
Using instrumental variable method, the third part of the dissertation investigates the effect of remittances on health capital accumulation. Total expenditure is divided into two categories: expenditure on medicines and expenditure on visits and laboratory services. The estimation is presented for two separate sub groups, rural and urban area.
The vector of covariates includes information related to individual and households characteristics such as; age, gender, schooling, area of residence etc. In the models, household incomes are considered separately from remittances in order to identify whether income from remittances has the same effect as other types of household non-labor income in the decision-making of household members.
The findings suggest a negative relationship between education and remittance receiving status. The estimation of the survival function indicates that receiving remittances from household members working abroad increases the hazard of leaving school after the end of secondary education. The negative relationship between education and remittances is more evident for males living in rural areas.
In line with previous studies, empirical results show that receiving remittances does not have any impact in the probability of working or hours worked per week by males. Receipt of remittances seems to impact the labor market behavior of females, because they reduce their hours worked in presence of remittances.
The last part of the study indicates that households increase their expenditure for medicines and other health services in the presence of remittance income. The positive relationship is statistically significant in the case of remittance receiving households living in the rural area. The magnitude is lower in the case of total expenditure for visits and laboratory. However, total expenditure for visits and laboratory are likely to have significant impact on the health outcome given their prevention nature.
These findings show that remittance flows pay a heterogeneous role in the decision making process of remittance-receiving household members. However, these non-labor income flows may play an important role in supporting expenditures, especially for those living in rural areas
On the use of Constructed Wetlands in mountain regions: innovative tools and configurations
No full textThe use of Constructed Wetlands (CWs) has been increasing over the last twenty years for decentralized wastewater treatment projects (e.g. rural communities, isolated houses, etc.) because of the low maintenance requirements and operational costs, efficiency in terms of organic matter, nitrogen and suspended solid removal. Nevertheless, the application of these systems in mountain areas is faced with some issues related to the specific characteristics of these areas, namely: the complex morphology with steep slopes and limited extensions of flat land, low temperatures and, in tourist contexts, population variations throughout the year. Limited availability of suitable land is a key issue for the application of a technology requiring considerable surfaces to produce effluents of good quality. Land area requirements constitute a well-known problem of CWs that is related to a lack of knowledge on the biological reactions occurring inside the bed. In fact, usually CWs are designed by considering simple first order decay models and specific surface area requirements, while the real requirements are not taken into account, leading most of the times to an overestimation of the area required. The limited knowledge on the processes and relative efficiencies of CW leads to overdesign of CW, mainly in low temperatures contexts and where there is a fluctuation on the resident population. Despite the efficiency that could be achieved through overestimation, those systems would be underutilized for a large part of the year. Ultimately, overestimated CWs consume more land than needed, eventually leading to the decision of switching to other systems.
This research aims to identify approaches and configurations that may improve the applicability of CWs for wastewater treatment of mountain communities. These approaches try to overcome the cross-cutting issue of land area requirement, as well as those related to the variation of temperature and population through the year. This was done by exploring the use of respirometric techniques for the estimation of kinetic and stoichiometric reactions inside the bed and by testing, in a pilot plant, the influence of the tourist presence and low temperatures on the efficiency of innovative CW configurations.
The research was developed at both the lab and the field scale. At the lab scale, two different tests were used in order to estimate the oxygen consumption in CW filter material: liquid respirometry and the off-gas technique. Liquid respirometry proved to be a reliable method when used to measure kinetic and stoichiometric parameters of the CW’s biomass. The off-gas technique was applied at the lab scale showing promising results, though further research is needed to improve the applicability of the method to CWs. Along with that, at the lab scale, a modified AUR method was applied on the CW material to quantify the nitrification rate of real systems at different temperatures and therefore to predict the removal efficiency throughout the year.
At the field scale, several tests were performed in a pilot plant composed by two hybrid CWs (VSSF+HSSF). Among these: operation under continuous and discontinuous winter conditions, operation with overload during the summer (to simulate the presence of tourists) and the application of innovative configurations (Recirculated and Aerated VSSF). All these tests were designed with the purpose of dealing with the trade-off between the reduction of a CW’s land area requirement and the enhancement of its efficiency. Two innovative configurations were tested in the pilot plant: Recirculated VSSF CW and Aerated VSSF CW. Both configurations can provide saturated and unsaturated conditions, which allow the nitrification/denitrification inside the bed. During the period when experimental configurations were tested, the traditional VSSF CW was operated with an average specific surface area to 3.5 m2/PE, the Recirculated VSSF of 1.5 m2/PE and the Aerated VSSF of 1.9 m2/PE on average. The results showed that the CW’s surface can be considerably reduced without a significant reduction in the removal efficiency. The extra investment needed to equip VSSF CWs with aeration/recirculation would be compensated by a lower area requirement.
This study explored some of the problems associated with the application of traditional CWs under the physical and social conditions that characterize mountain contexts, providing important information for future research and application. First of all, a reliable tool, the respirometric technique, was explored for the estimation of kinetic and stoichiometric parameters that will allow a more precise estimation of the land area required for these systems. Moreover, two innovative configurations (the use of recirculation and aeration in CWs) were proposed to be used where traditional configurations, though well designed, are still too large to be applied. Such configurations can also be used as a temporary solution to increase the treatment capacity during tourist peak seasons, while a traditional configuration is kept over the rest of the year. While this research focused on mountain environments, the configurations and results contained therein could be applied to a wide variety of settings where shortage of land or difficult climate conditions would exclude CWs from the list of wastewater treatment options available
Protein-dependent prediction of messenger RNA binding using Support Vector Machines
No full textRNA-binding proteins interact specifically with RNA strands to regulate important cellular processes. Knowing the binding partners of a protein is a crucial issue in biology and it is essential to understand the protein function and its involvement in diseases. The identification of the interactions is currently resolvable only through in vivo and in vitro experiments which may not detect all binding partners. Computational methods which capture the protein-dependent nature of the binding phenomena could help to predict, in silico, the binding and could be resistant against experimental biases.
This thesis addresses the creation of models based on support vector machines and trained on experimental data. The goal is the identification of RNAs which bind specifically to a regulatory protein. Starting from a case study, done with protein CELF1, we extend our approach and propose three methods to predict whether an RNA strand can be bound by a particular RNA-binding protein. The methods use support vector machines and different features based on the sequence (method Oli), the motif score (method OliMo) and the secondary structure (method OliMoSS). We apply them to different experimentally-derived datasets and compare the predictions with two methods: RNAcontext and RPISeq. Oli outperforms OliMoSS and RPISeq affirming our protein specific prediction and suggesting that oligo frequencies are good discriminative features. Oli and RNAcontext are the most competitive methods in terms of AUC. A Precision-Recall analysis reveals a better performance for Oli. On a second experimental dataset, where negative binding information is available, Oli outperforms RNAcontext with
a precision of 0.73 vs. 0.59. Our experiments show that features based on primary sequence information are highly discriminative to predict the binding between protein and RNA. Sequence motifs can improve the prediction only for some RNA-binding proteins. Finally, we can conclude that experimental data on RNA-binding can be effectively used to train protein-specific models for in silico predictions
Knowledge Based Open Entity Matching
No full textIn this work we argue for the definition a knowledge-based entity matching framework for the implementation of a reliable and incrementally scalable solution. Such knowledge base is formed by an ontology and a set of entity matching rules suitable to be applied as a reliable equational theory in the context of the Semantic Web. In particular, we are going to prove that relying on the existence of a set of contextual mappings to ease the semantic heterogeneity characterizing descriptions on the Web, a knowledge-based solution can perform comparably, and sometimes better, than existing solutions at the state of the art. We further argue that a knowledge-based solution to the open entity matching problem ought to be considered under the open world assumption, as in some cases the descriptions to be matched may not contain the necessary information to take any accurate matching decision. The main goal of this work is to show how the framework proposed is suitable to pursue a reliable solution of the entity matching problem, regardless the set of rules for the ontology adopted. In fact, we believe that structural and syntactic heterogeneity affecting data on the Web undermine the definition of a global unique solution. However, we argue that a knowledge-driven approach, considering the semantic and meta-properties of compared attributes, can provide important benefits and lead to more reliable solutions. To achieve this goal, we are going to implement several experiments to evaluate different sets of rules, testing our thesis and learning important lessons for future developments. The sets of rules that we will consider to bootstrap the solution proposed in this work are the result of diverse complementary processes: first we want to investigate whether capturing the matching knowledge employed by people in taking entity matching decision by relying on machine learning techniques can produce an effective set of rules (bottom-up strategy); second, we investigate the application of formal ontology pools to analyze the features defined in the ontology and support the definition of entity matching rules (top-down strategy). Moreover, in this work we argue that by merging the rules resulting from these complementary processes, we can define a set of rules that can support reliably entity matching decision in an open context