New kernel functions and learning methods for text and data mining

Recent advances in machine learning methods enable increasingly the automatic construction of various types of computer assisted methods that have been difficult or laborious to program by human experts. The tasks for which this kind of tools are needed arise in many areas, here especially in the fields of bioinformatics and natural language processing. The machine learning methods may not work satisfactorily if they are not appropriately tailored to the task in question. However, their learning performance can often be improved by taking advantage of deeper insight of the application domain or the learning problem at hand. This thesis considers developing kernel-based learning algorithms incorporating this kind of prior knowledge of the task in question in an advantageous way. Moreover, computationally efficient algorithms for training the learning machines for specific tasks are presented. In the context of kernel-based learning methods, the incorporation of prior knowledge is often done by designing appropriate kernel functions. Another well-known way is to develop cost functions that fit to the task under consideration. For disambiguation tasks in natural language, we develop kernel functions that take account of the positional information and the mutual similarities of words. It is shown that the use of this information significantly improves the disambiguation performance of the learning machine. Further, we design a new cost function that is better suitable for the task of information retrieval and for more general ranking problems than the cost functions designed for regression and classification. We also consider other applications of the kernel-based learning algorithms such as text categorization, and pattern recognition in differential display. We develop computationally efficient algorithms for training the considered learning machines with the proposed kernel functions. We also design a fast cross-validation algorithm for regularized least-squares type of learning algorithm. Further, an efficient version of the regularized least-squares algorithm that can be used together with the new cost function for preference learning and ranking tasks is proposed. In summary, we demonstrate that the incorporation of prior knowledge is possible and beneficial, and novel advanced kernels and cost functions can be used in algorithms efficiently.

The Figure of Radha in Miniature Paintings: From the Pastoral to the Courtly, from Text to Visuality, from Polyphony to Normativity

This article analyses how Radha was depicted in miniature paintings between the 16th and 19th century in North India. Interrogating the link between text and image, contrasting poetry, style and historical settings with the visual representations of this central figure, my reflections focus on the changing nature of Radha. Through various examples from miniature paintings of different periods and schools, this article analyses the way the rich personality of Radha was transposed into images. In order to stress the changes brought to this female figure, I compare her to Krishna, the masculine figure who is always at her side. The main goal of the article is to show the normative power of images on the figure of Radha, with normativity being understood as the simplification, iconisation, aestheticisation and stereotypification of a figure with polysemous references.

A Dependency Parsing Approach to Biomedical Text Mining

Biomedical research is currently facing a new type of challenge: an excess of information, both in terms of raw data from experiments and in the number of scientific publications describing their results. Mirroring the focus on data mining techniques to address the issues of structured data, there has recently been great interest in the development and application of text mining techniques to make more effective use of the knowledge contained in biomedical scientific publications, accessible only in the form of natural human language. This thesis describes research done in the broader scope of projects aiming to develop methods, tools and techniques for text mining tasks in general and for the biomedical domain in particular. The work described here involves more specifically the goal of extracting information from statements concerning relations of biomedical entities, such as protein-protein interactions. The approach taken is one using full parsing—syntactic analysis of the entire structure of sentences—and machine learning, aiming to develop reliable methods that can further be generalized to apply also to other domains. The five papers at the core of this thesis describe research on a number of distinct but related topics in text mining. In the first of these studies, we assessed the applicability of two popular general English parsers to biomedical text mining and, finding their performance limited, identified several specific challenges to accurate parsing of domain text. In a follow-up study focusing on parsing issues related to specialized domain terminology, we evaluated three lexical adaptation methods. We found that the accurate resolution of unknown words can considerably improve parsing performance and introduced a domain-adapted parser that reduced the error rate of theoriginal by 10% while also roughly halving parsing time. To establish the relative merits of parsers that differ in the applied formalisms and the representation given to their syntactic analyses, we have also developed evaluation methodology, considering different approaches to establishing comparable dependency-based evaluation results. We introduced a methodology for creating highly accurate conversions between different parse representations, demonstrating the feasibility of unification of idiverse syntactic schemes under a shared, application-oriented representation. In addition to allowing formalism-neutral evaluation, we argue that such unification can also increase the value of parsers for domain text mining. As a further step in this direction, we analysed the characteristics of publicly available biomedical corpora annotated for protein-protein interactions and created tools for converting them into a shared form, thus contributing also to the unification of text mining resources. The introduced unified corpora allowed us to perform a task-oriented comparative evaluation of biomedical text mining corpora. This evaluation established clear limits on the comparability of results for text mining methods evaluated on different resources, prompting further efforts toward standardization. To support this and other research, we have also designed and annotated BioInfer, the first domain corpus of its size combining annotation of syntax and biomedical entities with a detailed annotation of their relationships. The corpus represents a major design and development effort of the research group, with manual annotation that identifies over 6000 entities, 2500 relationships and 28,000 syntactic dependencies in 1100 sentences. In addition to combining these key annotations for a single set of sentences, BioInfer was also the first domain resource to introduce a representation of entity relations that is supported by ontologies and able to capture complex, structured relationships. Part I of this thesis presents a summary of this research in the broader context of a text mining system, and Part II contains reprints of the five included publications.

Diagnosis of hepatic steatosis by contrast-enhanced abdominal computed tomography

Objective To evaluate the diagnostic capacity of abdominal computed tomography in the assessment of hepatic steatosis using the portal phase with a simplified calculation method as compared with the non-contrast-enhanced phase. Materials and Methods In the present study, 150 patients were retrospectively evaluated by means of non-contrast-enhanced and contrast-enhanced computed tomography. One hundred patients had hepatic steatosis and 50 were control subjects. For the diagnosis of hepatic steatosis in the portal phase, the authors considered a result of < 104 HU calculated by the formula [L - 0.3 × (0.75 × P + 0.25 × A)] / 0.7, where L, P and A represent the attenuation of the liver, of the main portal vein and abdominal aorta, respectively. Sensitivity, specificity, positive and negative predictive values were calculated, using non-contrast-enhanced computed tomography as the reference standard. Results The simplified calculation method with portal phase for the diagnosis of hepatic steatosis showed 100% sensitivity, 36% specificity, negative predictive value of 100% and positive predictive value of 75.8%. The rate of false positive results was 64%. False negative results were not observed. Conclusion The portal phase presents an excellent sensitivity in the diagnosis of hepatic steatosis, as compared with the non-contrast-enhanced phase of abdominal computed tomography. However, the method has low specificity.

Construction of an index of information from clinical practice in Radiology and Imaging Diagnosis based on text mining and thesaurus

Objective To construct a Portuguese language index of information on the practice of diagnostic radiology in order to improve the standardization of the medical language and terminology. Materials and Methods A total of 61,461 definitive reports were collected from the database of the Radiology Information System at Hospital das Clínicas – Faculdade de Medicina de Ribeirão Preto (RIS/HCFMRP) as follows: 30,000 chest x-ray reports; 27,000 mammography reports; and 4,461 thyroid ultrasonography reports. The text mining technique was applied for the selection of terms, and the ANSI/NISO Z39.19-2005 standard was utilized to construct the index based on a thesaurus structure. The system was created in *html. Results The text mining resulted in a set of 358,236 (n = 100%) words. Out of this total, 76,347 (n = 21%) terms were selected to form the index. Such terms refer to anatomical pathology description, imaging techniques, equipment, type of study and some other composite terms. The index system was developed with 78,538 *html web pages. Conclusion The utilization of text mining on a radiological reports database has allowed the construction of a lexical system in Portuguese language consistent with the clinical practice in Radiology.

Text i context. El concurs de cinema amateur «JULIUS» (1987-1996)

El «Julius» és un concurs de cinema amateur que fonamenta la seva singularitat en el fet de partir d’un text literari prefixat del qual els participants han de fer una adaptació audiovisual. En aquest article s’estudia la primera època del concurs: el seu origen, les diverses edicions en què es dugué a terme, les incidències que s’hi produïren..., i el context social i cultural que va contribuir a fer-ne un concurs amb unes característiques úniques que es concretaren en l’anomenat «esperit Julius», sorneguer, llibertari i surrealista.

Magnetic resonance imaging and BMB score in the evaluation of bone involvement in Gaucher’s disease patients

AbstractObjective:To evaluate by magnetic resonance imaging changes in bone marrow of patients undergoing treatment for type I Gaucher’s disease.Materials and Methods:Descriptive, cross-sectional study of Gaucher’s disease patients submitted to 3 T magnetic resonance imaging of femurs and lumbar spine. The images were blindly reviewed and the findings were classified according to the semiquantitative bone marrow burden (BMB) scoring system.Results:All of the seven evaluated patients (three men and four women) presented signs of bone marrow infiltration. Osteonecrosis of the femoral head was found in three patients, Erlenmeyer flask deformity in five, and no patient had vertebral body collapse. The mean BMB score was 11, ranging from 9 to 14.Conclusion:Magnetic resonance imaging is currently the method of choice for assessing bone involvement in Gaucher’s disease in adults due to its high sensitivity to detect both focal and diffuse bone marrow changes, and the BMB score is a simplified method for semiquantitative analysis, without depending on advanced sequences or sophisticated hardware, allowing for the classification of the disease extent and assisting in the treatment monitoring.

Jane Eyre: del text a la pantalla : Un estudi comparatiu entre l'obra de Charlotte Brontë i l'adaptació cinematogràfica de Cary Fukunaga

Treball de fi de màster que estudia comparativament les relacions entre la novel·la "Jane Eyre" de Charlotte Brontë i l'adaptació cinematogràfica de la mateixa realitzada pel director Cary Fukunaga.

Modelo teórico quântico para o processo de adsorção física

This article introduces a simplified model for the theoretical study of the physical adsorption process of gaseous He on the planes (100) and (111) of the solid Xe matrix, whose crystalline structure is face centered cubic (fcc). The Ab initio calculations were carried out at the MP2 level of theory employing basis sets obtained through the Generator Coordinate Method, where the core electrons were represented by a pseudopotential. The calculated adsorption energies for the (100) and (111) faces are 5,39 and 4,18 kJ/mol, respectively. This simplified model is expected to be suitable for treating complex systems of applied interest.

Abordagens em síntese assimétrica

The approaches in asymmetric synthesis as the chiron approach, chiral auxiliaries, chiral reagents and asymmetric catalysis are described in a simplified way.