Este trabalho de doutorado apresenta duas inovações: a Discriminative Paraconsistent Machine (DPM), que consiste em um novo classificador elaborado com base na lógica paraconsistente anotada (LPA) e a aplicação da DPM para a identificação de patologias na laringe, por meio de exames nos sinais de voz de um locutor. Não há relatos na literatura sobre o uso da LPA para construção de um classificador e sobre suas aplicações para a finalidade proposta. Os resultados obtidos são motivadores, indicando um avanço na área. / This PhD thesis presents two novelties: the Discriminative Paraconsistent Machine (DPM), which is a new classifier built on the basis of the annotated paraconsistent logic (APL), and the applications of DPM to identify larynx pathologies, by inspecting a voice signal. There is neither a comment on literature about the use of APL to built a classifier nor about its applications for the proposed application. The results obtained create motivation, showing a clear progress in the field.

Acoustic modeling in state-of-the-art speech recognition systems is commonly based on discriminative criteria. Different from the paradigm of the conventional distribution estimation such as maximum a posteriori (MAP) and maximum likelihood (ML), the most popular discriminative criteria such as MCE and MPE aim at direct minimization of the empirical error rate. As recent ASR applications become diverse, it has been increasingly recognized that realistic applications often require a model that can be optimized for a task-specific goal or a particular scenario beyond the general purposes of the current discriminative criteria. These specific requirements cannot be directly handled by the current discriminative criteria since the objective of the criteria is to minimize the overall empirical error rate. In this thesis, we propose novel objective-driven discriminative training and adaptation frameworks, which are generalized from the minimum classification error (MCE) criterion, for various tasks and scenarios of speech recognition and detection. The proposed frameworks are constructed to formulate new discriminative criteria which satisfy various requirements of the recent ASR applications. In this thesis, each objective required by an application or a developer is directly embedded into the learning criterion. Then, the objective-driven discriminative criterion is used to optimize an acoustic model in order to achieve the required objective. Three task-specific requirements that the recent ASR applications often require in practice are mainly taken into account in developing the objective-driven discriminative criteria. First, an issue of individual error minimization of speech recognition is addressed and we propose a direct minimization algorithm for each error type of speech recognition. Second, a rapid adaptation scenario is embedded into formulating discriminative linear transforms under the MCE criterion. A regularized MCE criterion is proposed to efficiently improve the generalization capability of the MCE estimate in a rapid adaptation scenario. Finally, the particular operating scenario that requires a system model optimized at a given specific operating point is discussed over the conventional receiver operating characteristic (ROC) optimization. A constrained discriminative training algorithm which can directly optimize a system model for any particular operating need is proposed. For each of the developed algorithms, we provide an analytical solution and an appropriate optimization procedure.

Nitrous oxide (N2O) is a widely used anesthetic adjunct in dentistry and medicine that is also commonly abused. N2O alters the function of several receptors in vitro and ex vivo, however, the receptors systems underlying its abuse-related intoxicating effects are poorly understood. The goals of this dissertation were to (1) establish N2O as a discriminative stimulus, (2) characterize the temporal properties of the discriminative stimulus, (3) determine the degree of similarity between N2O and other inhalants and (4) explore the neurochemical effects responsible for the stimulus properties of N2O. Twenty-four mice were trained to discriminate 10 minutes exposure to 60% N2O+40% O2 from 100% O2 in daily 5 minute food-reinforced operant sessions. Mice acquired the discrimination in a mean of 38 sessions. N2O produced concentration-dependent full substitution for itself. Full substitution required 7 minutes of N2O exposure but the offset of stimulus effects following cessation of N2O exposure were more rapid. Varying degrees of partial substitution for N2O were engendered by abused vapors and vapor anesthetics. The aromatic hydrocarbon toluene produced the most robust substitution for N2O. One or more toluene concentrations produced full substitution for N2O in 7 of 8 subjects, suggesting that these two abused inhalants share common neurochemical mechanisms. The NMDA receptor open channel blockers (+)-MK-801, ketamine and memantine produced dose-dependent partial substitution for N2O. A competitive NMDA antagonist and NMDA glycine site antagonist did not substitute for N2O. Pretreatment with (+)-MK-801 as well as ethanol produced dose-dependent leftward shifts in the N2O concentration effect curve further suggesting some overlap in their mechanisms of action. GABAA agonists and positive allosteric modulators, opioid agonists, serotonergic agonists, nicotine, a nNOS inhibitor and the psychomotor stimulant amphetamine all failed to appreciably substitute for N2O and/or failed to alter the N2O concentration effect curve when administered prior to N2O exposure. No drug tested produced greater than 80% mean N2O-lever selection leaving open the possibility of other neurochemical contributors to the stimulus effects of N2O.

In this work, an extensible word-alignment framework is implemented from scratch. It is based on a discriminative method that combines a wide range of lexical association measures and other features and requires a small amount of manually word-aligned data to optimize parameters of the model. The optimal alignment is found as minimum-weight edge cover, selected suboptimal alignments are used to estimate confidence of each alignment link. Feature combination is tuned in the course of many experiments with respect to the results of evaluation. The evaluation results are compared to GIZA++. The best trained model is used to word-align a large Czech-English parallel corpus and from the links of highest confidence a bilingual lexicon is extracted. Single-word translation equivalents are sorted by their significance. Lexicons of different sizes are extracted by taking top N translations. Precision of the lexicons is evaluated automatically and also manually by judging random samples.

This thesis is compromised of three studies in which basic principles of operant conditioning were applied to horses. Autoshaping was examined as a method for horse training. Observational learning was investigated to confirm that naive horses can, in fact, acquire novel behavior by observing experienced horses, and the rate of acquisition with observation is more rapid than spontaneous responding without observation. A third study examined the effect of discriminative stimulus intensity on the acquisition rate of novel behavior. All subjects learned to use an operant conditioning device. Subjects in the first study autoshaped. Observational learning was also demonstrated to be a means by which horses can learn. The rate of learning was significantly improved through observation. Intensity of the discriminative stimulus affects the acquisition of novel behavior. The subject exposed to the higher intensity stimulus acquired sustained manipulandum pressing significantly faster than other subjects. It was concluded that horses acquire behavior in much the same manner as other species.

[Truncated abstract] Some children with Specific Language Impairment (SLI) show poor performance on behavioural tasks designed to measure rapid auditory processing, such as the Repetition Test developed by Tallal and colleagues. Stemming from concerns about whether this task reflects higher-order, cognitive variables, this thesis sought to determine whether the performance deficits SLI children show were evident at the neurophysiological level, with minimal cognitive influences, such as attention, using paradigms designed to elicit the mismatch negativity (MMN) and late discriminative negativity (LDN). In the first two studies, a MMN paradigm, equivalent to the Repetition Test, was trialled with a group of 8 adults. In this paired paradigm, the second tone of a pair of pure tones ascending in frequency ('low'-'high') was occasionally replaced with a 'low' tone. The aim was to determine a 'long' and 'short' intra-pair interval (IPI) with which MMN was generated utilizing this paradigm and that were congruent with the Repetition Test findings (i.e., a long IPI at which SLI were able to perform the task, and a short IPI at which SLI children's performance was selectively impaired). In Study One, MMN to a within-pair frequency change was generated with the 30 ms IPI, but not the 700 ms IPI. The grouping parameters of the temporal window of integration (TWI) and temporal distinctiveness were considered less than optimal for the grouping of the pairs presented at 700 ms IPI for the pre-attentive system to register the within-pair frequency change. ... The frequency difference limens (DLs) of the SLI group were significantly higher than a group of 18 normally developing age- and intelligence-matched peers, but there was no significant difference between the groups in their performance on a control intensity discrimination task. The iii SLI group also showed poorer reading skills, yet frequency discrimination was related to oral language ability only. In the final study, MMN was measured to examine the pre-attentive neurophysiological basis of the SLI group's frequency discrimination deficit. Two frequency deviants that were just above each group's 75% DL on the frequency discrimination task were employed in a simple frequency change paradigm: 40 Hz difference for the control group, and 80 Hz difference for the SLI group. MMN and LDN were elicited in the group of 15 normally developing children to their 40 Hz suprathreshold frequency difference and to the 80 Hz difference. A significant MMN was not observed in the group of 13 SLI children to the 40 Hz difference, which was below their threshold level. However, despite discrimination at the behavioural level, MMN did not reach significance in the SLI group to their 80 Hz suprathreshold frequency difference, yet LDN was observed. MMN was larger in both groups for the 80 Hz difference. Furthermore, MMN and LDN amplitude to the suprathreshold deviants were predictive of both frequency and intensity DLs. These results suggested that SLI children have a pre-attentive neurophysiologically mediated insensitivity to small frequency differences, and that MMN (and LDN) to suprathreshold frequency deviants is a sensitive indicator of group discrimination differences and brain-behaviour relationships in children with and without SLI.

This thesis addresses the problems of phonemic variability and confusability from the pronunciation modeling perspective for an automatic speech recognition (ASR) system. In particular, several research directions are investigated. First, automatic grapheme-to- phoneme (g2p) and phoneme-to-phoneme (p2p) converters are developed that generate alternative pronunciations for in-vocabulary as well as out-of-vocabulary (OOV) terms. Since the addition of alternative pronunciation may introduce homophones (or close homophones), there is an increase of the confusability of the system. A novel measure of this confusability is proposed to analyze it and study its relation with the ASR performance. This pronunciation confusability is higher if pronunciation probabilities are not provided and can potentially severely degrade the ASR performance. It should, thus, be taken into account during pronunciation generation. Discriminative training approaches are, then, investigated to train the weights of a phoneme confusion model that allows alternative ways of pronouncing a term counterbalancing the phonemic confusability problem. The objective function to optimize is chosen to correspond to the performance measure of the particular task. In this thesis, two tasks are investigated, the ASR task and the KeywordSpotting (KWS) task. For ASR, an objective that minimizes the phoneme error rate is adopted. For experiments conducted on KWS, the Figure of Merit (FOM), a KWS performance measure, is directly maximized.

This essay is about the somatosensory system, which is divided into different kinds of touch. Described briefly are the proprioceptive touch, which is transported to the brain via A-alfa fibers and transmits information about e. g. limb position and movement. The cutaneous touch is the main focus and it is divided into discriminative touch and affective touch. The first corresponds to stimuli such as vibration and pressure and is transported via A-beta axons. The second, affective touch, corresponds to e.g. painful and pleasant stimuli which are transported to the brain via A-delta and C-fibers. The aim of the essay is to give an overview of the sense of touch, by doing a literature search, including a discussion of relevant neuronal correlates focusing particularly on affective touch. Moreover, the physiological aspects of touch will be presented. The sources that are used are review and original articles taken from databases such as ScienceDirect, and some articles send by the author. Some books have also been used to find more general knowledge. The conclusion for the essay is that touch is important for humans to function in everyday life. Additional, a specific receptor called C- tactile (CT) is identified to correspond to gentle touch and is suggested to have a vital role for humans in maintaining and forming social bounds. Moreover, discriminative touch is associated with activation in the primary and secondary somatosensory cortex, whereas affective touch seems to be associated with activity in the orbitofrontal cortex, cingulate cortex and the insula cortex, as well as the prefrontal cortex, which is suggested to be activated during interpersonal touch. Further, the sense touch needs to be more researched in order to understand its functions and benefits deeper.

This thesis presents a commercial outcome prediction system (CPS) capable of predicting the likely future monetary return that would be generated by an invention. The CPS is designed to be used by university technology transfer offices for invention assessment purposes, and is based on the data from their historical invention cases. It is aimed at improving technology transfer offices' invention assessment performance. Using qualitative critical factors suggested by literature. a prototype CPS based on decision tree induction was developed. The prediction performance achieved by the prototype CPS was unreliable. Three surveys with various technology transfer offices were then performed, and the findings were incorporated into a final version of the CPS, which was based on neural networks. Subject to information obtained in the surveys, a number of potentially predictive attributes were proposed to form part of the predictor variables for the CPS. The CPS starts with a number of data reduction operations (based on principal component analysis and decision tree techniques), which identify the critical predictor variables. The CPS then uses five neural-network training algorithms to generate candidate classifiers, upon which the final classification is based. The prediction results achieved by the CPS were good and reliable. Additionally, the data reduction operations successfully captured the most discriminative invention attributes. The research demonstrated the potential or using the CPS for invention assessment. However, it requires sufficient historical data from the technology transfer office using it to provide accurate assessments.

Research Doctorate - Doctor of Philosophy (PhD) / This thesis investigates mechanisms of brain maturation by utilising the special advantages offered by the protracted maturation of neural circuits in chicken forebrain. Biochemical, behavioural and electrophysiological techniques are used in behaving animals to investigate the functional consequences of maturation changes at the molecular, behavioural and physiological levels. Two issues are addressed: (1) do immature (2 week) and mature (8 week) chickens employ different molecular mechanisms to produce changes in neuronal function after learning a behavioural task; and (2) can quantitative non-invasive measures of neuronal function be used to monitor maturation changes in chicken forebrain? Biochemical investigation of subcellular fractions using antibodies and western blots of chicken forebrain and intermediate medial mesopallium (IMM) revealed regional differences in expression levels of a number of components of the glutamatergic neurotransmitter system. The discriminative taste aversion learning (DTAL) task was used to assess whether an animal learns the same task at different ages using different intracellular signalling pathways. The patterns of biochemical change seen in the IMM after DTAL training was very different at 2 weeks and 8 weeks. Two major differences were observed. Firstly, the same type of training induced changes occurred at both ages in GluR1 and CaMKII but they occurred faster at 8 weeks. Secondly the difference in ERK and CREB responses is consistent with a change in the relative contribution made by the ERK signalling pathway and CREB requirement to learning at these two ages. These results imply that the molecular changes induced by learning a behavioural task are faster in mature than immature brain and may involve a different balance of intracellular signalling pathways. In order to be able to investigate biological mechanisms controlling maturation and to use the chicken as an animal model in which pharmacological and/or environmental agents can be screened for potentially harmful effects on brain maturation two non-invasive measures of neuronal function were investigated. One was behavioural (prepulse inhibition: PPI) and the other was electrophysiological (auditory evoked related potentials: AERP). PPI in the chicken was examined electromyographically and via whole body movement with a stabilimeter apparatus. In two strains of chicken (a meat breed and a laying breed) PPI was identified but shown to be small and variable compared to that in the rat. The results indicate that the phenomenon of PPI in the chicken is too small and variable to be used as a quantitative measure of neural circuit maturation. Quantitative analysis of the chicken AERP revealed a significant decrease in amplitude of the positive AERP component and a decrease in latency of the negative AERP component with maturation. These maturation changes were comparable to developmental changes seen in human and other mammal AERPs. Such changes may reflect changes in the intracortical synaptic organisation of the auditory cortex. This technique allowed for repeated measures to be undertaken on the same animal over a number of weeks and enabled developmental changes to be monitored. This technique was extended to investigate perturbed maturation via the induction of chemically induced hypothyroidism. Results from this study showed that the induction of late onset hypothyroidism produces measurable effects on the chicken AERP consistent with perturbation in maturation of neuronal circuits and synapses. This suggests that AERPs may be useful non-invasive functional measures of brain maturation that can be used to study the effects of endogenous or exogenous factors on brain maturation in the chicken. Since human brain also exhibits a protracted maturation period the availability of a well characterised animal model for protracted brain maturation provides an opportunity to identify molecules, genes and environmental factors that are important in the regulation of maturation. Such a model may provide the basis for developing rational therapies or prevention strategies for some neurodevelopmental disorders. The protracted maturation of neuronal circuits observed in chicken forebrain offers such a model.