Apprentissage automatique de la morphologie : le cas des structures racine-schème

RÉSUMÉ Cette thèse porte sur le développement de méthodes algorithmiques pour découvrir automatiquement la structure morphologique des mots d'un corpus. On considère en particulier le cas des langues s'approchant du type introflexionnel, comme l'arabe ou l'hébreu. La tradition linguistique décrit la morphologie de ces langues en termes d'unités discontinues : les racines consonantiques et les schèmes vocaliques. Ce genre de structure constitue un défi pour les systèmes actuels d'apprentissage automatique, qui opèrent généralement avec des unités continues. La stratégie adoptée ici consiste à traiter le problème comme une séquence de deux sous-problèmes. Le premier est d'ordre phonologique : il s'agit de diviser les symboles (phonèmes, lettres) du corpus en deux groupes correspondant autant que possible aux consonnes et voyelles phonétiques. Le second est de nature morphologique et repose sur les résultats du premier : il s'agit d'établir l'inventaire des racines et schèmes du corpus et de déterminer leurs règles de combinaison. On examine la portée et les limites d'une approche basée sur deux hypothèses : (i) la distinction entre consonnes et voyelles peut être inférée sur la base de leur tendance à alterner dans la chaîne parlée; (ii) les racines et les schèmes peuvent être identifiés respectivement aux séquences de consonnes et voyelles découvertes précédemment. L'algorithme proposé utilise une méthode purement distributionnelle pour partitionner les symboles du corpus. Puis il applique des principes analogiques pour identifier un ensemble de candidats sérieux au titre de racine ou de schème, et pour élargir progressivement cet ensemble. Cette extension est soumise à une procédure d'évaluation basée sur le principe de la longueur de description minimale, dans- l'esprit de LINGUISTICA (Goldsmith, 2001). L'algorithme est implémenté sous la forme d'un programme informatique nommé ARABICA, et évalué sur un corpus de noms arabes, du point de vue de sa capacité à décrire le système du pluriel. Cette étude montre que des structures linguistiques complexes peuvent être découvertes en ne faisant qu'un minimum d'hypothèses a priori sur les phénomènes considérés. Elle illustre la synergie possible entre des mécanismes d'apprentissage portant sur des niveaux de description linguistique distincts, et cherche à déterminer quand et pourquoi cette coopération échoue. Elle conclut que la tension entre l'universalité de la distinction consonnes-voyelles et la spécificité de la structuration racine-schème est cruciale pour expliquer les forces et les faiblesses d'une telle approche. ABSTRACT This dissertation is concerned with the development of algorithmic methods for the unsupervised learning of natural language morphology, using a symbolically transcribed wordlist. It focuses on the case of languages approaching the introflectional type, such as Arabic or Hebrew. The morphology of such languages is traditionally described in terms of discontinuous units: consonantal roots and vocalic patterns. Inferring this kind of structure is a challenging task for current unsupervised learning systems, which generally operate with continuous units. In this study, the problem of learning root-and-pattern morphology is divided into a phonological and a morphological subproblem. The phonological component of the analysis seeks to partition the symbols of a corpus (phonemes, letters) into two subsets that correspond well with the phonetic definition of consonants and vowels; building around this result, the morphological component attempts to establish the list of roots and patterns in the corpus, and to infer the rules that govern their combinations. We assess the extent to which this can be done on the basis of two hypotheses: (i) the distinction between consonants and vowels can be learned by observing their tendency to alternate in speech; (ii) roots and patterns can be identified as sequences of the previously discovered consonants and vowels respectively. The proposed algorithm uses a purely distributional method for partitioning symbols. Then it applies analogical principles to identify a preliminary set of reliable roots and patterns, and gradually enlarge it. This extension process is guided by an evaluation procedure based on the minimum description length principle, in line with the approach to morphological learning embodied in LINGUISTICA (Goldsmith, 2001). The algorithm is implemented as a computer program named ARABICA; it is evaluated with regard to its ability to account for the system of plural formation in a corpus of Arabic nouns. This thesis shows that complex linguistic structures can be discovered without recourse to a rich set of a priori hypotheses about the phenomena under consideration. It illustrates the possible synergy between learning mechanisms operating at distinct levels of linguistic description, and attempts to determine where and why such a cooperation fails. It concludes that the tension between the universality of the consonant-vowel distinction and the specificity of root-and-pattern structure is crucial for understanding the advantages and weaknesses of this approach.

Bilateral Sequential NAION: A Comparison of Visual Outcomes in Fellow Eyes Using Quantitative Analysis of Goldmann Visual Fields

Purpose: Previous studies of the visual outcome in bilateral non-arteritic anterior ischemic optic neuropathy (NAION) have yielded conflicting results, specifically regarding congruity between fellow eyes. Prior studies have used measures of acuity and computerized perimetry but none has compared Goldmann visual field outcomes between fellow eyes. In order to better define the concordance of visual loss in this condition, we reviewed our cases of bilateral sequential NAION, including measures of visual acuity, pupillary function and both pattern and severity of visual field loss.Methods: We performed a retrospective chart review of 102 patients with a diagnosis of bilateral sequential NAION. Of the 102 patients, 86 were included in the study for analysis of final visual outcome between the affected eyes. Visual function was assessed using visual acuity, Goldmann visual fields, color vision and RAPD. A quantitative total visual field score and score per quadrant was analyzed for each eye using the numerical Goldmann visual field scoring method previously described by Esterman and colleagues. Based upon these scores, we calculated the total deviation and pattern deviation between fellow eyes and between eyes of different patients. Statistical significance was determined using nonparametric tests.Results: A statistically significant correlation was found between fellow eyes for multiple parameters, including logMAR visual acuity (P = 0.0101), global visual field (P = 0.0001), superior visual field (P = 0.0001), and inferior visual field (P = 0.0001). In addition, the mean deviation of both total (P = 0.0000000007) and pattern (P = 0.000000004) deviation analyses was significantly less between fellow eyes ("intra"-eyes) than between eyes of different patients ("inter"-eyes).Conclusions: Visual function between fellow eyes showed a fair to moderate correlation that was statistically significant. The pattern of vision loss was also more similar in fellow eyes than between eyes of different patients. These results may help allow better prediction of visual outcome for the second eye in patients with NAION. These findings may also be useful for evaluating efficacy of therapeutic interventions.

Effect of a 1-hour single bout of moderate-intensity exercise on fat oxidation kinetics.

The present study aimed to examine the effects of a prior 1-hour continuous exercise bout (CONT) at an intensity (Fat(max)) that elicits the maximal fat oxidation (MFO) on the fat oxidation kinetics during a subsequent submaximal incremental test (IncrC). Twenty moderately trained subjects (9 men and 11 women) performed a graded test on a treadmill (Incr), with 3-minute stages and 1-km.h(-1) increments. Fat oxidation was measured using indirect calorimetry and plotted as a function of exercise intensity. A mathematical model (SIN) including 3 independent variables (dilatation, symmetry, and translation) was used to characterize the shape of fat oxidation kinetics and to determine Fat(max) and MFO. On a second visit, the subjects performed CONT at Fat(max) followed by IncrC. After CONT performed at 57% +/- 3% (means +/- SE) maximal oxygen uptake (Vo(2max)), the respiratory exchange ratio during IncrC was lower at every stage compared with Incr (P < .05). Fat(max) (56.4% +/- 2.3% vs 51.5% +/- 2.4% Vo(2max), P = .013), MFO (0.50 +/- 0.03 vs 0.40 +/- 0.03 g.min(-1), P < .001), and fat oxidation rates from 35% to 70% Vo(2max) (P < .05) were significantly greater during IncrC compared with Incr. However, dilatation and translation were not significantly different (P > .05), whereas symmetry tended to be greater in IncrC (P = .096). This study showed that the prior 1-hour continuous moderate-intensity exercise bout increased Fat(max), MFO, and fat oxidation rates over a wide range of intensities during the postexercise incremental test. Moreover, the shape of the postexercise fat oxidation kinetics tended to have a rightward asymmetry.

A Variscan pressure-temperature-time path for the N-E Mont Blanc massif

The northeastern portion of the Mont Blanc massif in western Switzerland is predominantly comprised of the granitic rocks of the Mont Blanc intrusive suit, and the Mont Blanc basement gneisses. Within these metamorphic rocks are a variety of sub-economic Fe skarns. The mineral assemblages and fluid inclusions from these rocks have been used to derive age, pressure, temperature and fluid composition constraints for two Variscan events. Metamorphic hornblendes within the assemblages from the basement amphibolites and iron sk:lms have been dated using Ar-40/Ar-39, and indicate that these metamorphic events have a minimum age of approximately 334 Ma. Garnet-hornblende-plagioclase thermobarometry and stable isotope data obtained from the basement amphibolites are consistent with metamorphic temperatures in the range 515 to 580 degrees C, and pressures ranging from 5 to 8 kbar. Garnet-hornblende-magnetite thermobarometry and fluid inclusion studies indicate that the iron skarns formed at slightly lower temperatures, ranging from 400 to 500 degrees C in the presence of saline fluids at formational pressures similar to those experienced by the basement amphibolites. Late Paleozoic minimum uplift rates and geothermal gradients calculated using these data and the presence of Ladinien ichnofossils are on the order of 0.32 mm/year and 20 degrees C/km respectively. These uplift rates and geothermal gradients differ from those obtained from the neighbouring Aiguilles Rouges massif and indicate that these two massifs experienced different metamorphic conditions during the Carboniferous and Permian periods. During the early to late Carboniferous period the relative depths of the two massifs were reversed with the Aiguilles Rouges being initially unroofed at a much greater rate than the Mont Blanc, but experiencing relatively slower uplift rates near the termination of the Variscan orogeny.

The data treatment dependent variability of U-Pb zircon ages obtained using mono-collector, sector field, laser ablation ICP-MS

U-Pb dating of zircons by laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) is a widely used analytical technique in Earth Sciences. For U-Pb ages below 1 billion years (1 Ga), Pb-206/U-238 dates are usually used, showing the least bias by external parameters such as the presence of initial lead and its isotopic composition in the analysed mineral. Precision and accuracy of the Pb/U ratio are thus of highest importance in LA-ICPMS geochronology. We consider the evaluation of the statistical distribution of the sweep intensities based on goodness-of-fit tests in order to find a model probability distribution fitting the data to apply an appropriate formulation for the standard deviation. We then discuss three main methods to calculate the Pb/U intensity ratio and its uncertainty in the LA-ICPMS: (1) ratio-of-the-mean intensities method, (2) mean-of-the-intensity-ratios method and (3) intercept method. These methods apply different functions to the same raw intensity vs. time data to calculate the mean Pb/U intensity ratio. Thus, the calculated intensity ratio and its uncertainty depend on the method applied. We demonstrate that the accuracy and, conditionally, the precision of the ratio-of-the-mean intensities method are invariant to the intensity fluctuations and averaging related to the dwell time selection and off-line data transformation (averaging of several sweeps); we present a statistical approach how to calculate the uncertainty of this method for transient signals. We also show that the accuracy of methods (2) and (3) is influenced by the intensity fluctuations and averaging, and the extent of this influence can amount to tens of percentage points; we show that the uncertainty of these methods also depends on how the signal is averaged. Each of the above methods imposes requirements to the instrumentation. The ratio-of-the-mean intensities method is sufficiently accurate provided the laser induced fractionation between the beginning and the end of the signal is kept low and linear. We show, based on a comprehensive series of analyses with different ablation pit sizes, energy densities and repetition rates for a 193 nm ns-ablation system that such a fractionation behaviour requires using a low ablation speed (low energy density and low repetition rate). Overall, we conclude that the ratio-of-the-mean intensities method combined with low sampling rates is the most mathematically accurate among the existing data treatment methods for U-Pb zircon dating by sensitive sector field ICPMS.

L'autre du même : emprunts et répétitions dans le "Roman de Perceforest"

Issu de la fin du Moyen Âge, le Roman de Perceforest est le plus long texte que le Moyen Âge nous ait laissé. Il entend décrire la vie des ancêtres préchrétiens d'Arthur et de ses chevaliers en les faisant descendre d'Alexandre le Grand. Au fil de son récit, l'auteur met en place une véritable poétique de la reprise, tant externe qu'interne. Il multiplie les références à des textes préexistants issus de différentes matières et va même jusqu'à en intégrer des morceaux entiers. Il reprend par ailleurs des parties de sa propre oeuvre, n'hésitant pas à reproduire des schémas, voire des séquences narratives précises. Cette esthétique répétitive a plusieurs conséquences, tant sur l'appartenance générique du texte, sur sa construction, que sur sa réception par le lecteur. Elle est au coeur de notre étude. Nous nous intéressons plus spécifiquement à un phénomène particulier de reprise qui montre l'intégration d'une séquence préexistante dans un contexte distinct de son apparition d'origine et que nous qualifions d'emprunt. Notre travail s'organise autour d'un examen successif des différents types d'emprunts qui apparaissent au sein du texte, tant intertextuels qu'intratextuels. À terme, c'est la cohérence et l'individualité du Perceforest, ainsi que la conception de l'écriture qui anime son auteur que nous mettons en avant, tout en proposant des jalons pour une théorie générale de l'emprunt. - The Roman de Perceforest, the longest known text from the Middle Ages, aims to describe the life of Arthur's pre-christian ancestors and knights, presenting them as descendants of Alexander the Great. Along the storytelling, a genuine poetics of external and internal repetition takes place: the Perceforest s authors multiply the references to previous texts belonging to several materials, integrating sometimes entire parts of other texts. Furthermore the narrative reproduces its own patterns and particular sequences in different places of the text. Throughout this research, we consider the influence of such repetition aesthetics on the literary genre definition, on its construction as well as on the reader's reception. In this dissertation, we explore a specific category of repetitions where pre-existing sequences are embedded in a narrative context that differs from the original context of occurrence, which we called emprunt (s). Reviewing the species of inter- and intra-textual recurrences occurring within the text, we reveal some overlooked aspects of the consistency, the specificity of the Perceforest and its author's idea of writing, striving to groundwork on general theory of «emprunts» that shall thereby be laid.

Effects of 2 different prior endurance exercises on whole-body fat oxidation kinetics: light vs. heavy exercise.

This study aimed to compare the effects of 2 different prior endurance exercises on subsequent whole-body fat oxidation kinetics. Fifteen men performed 2 identical submaximal incremental tests (Incr2) on a cycle ergometer after (i) a ∼40-min submaximal incremental test (Incr1) followed by a 90-min continuous exercise performed at 50% of maximal aerobic power-output and a 1-h rest period (Heavy); and (ii) Incr1 followed by a 2.5-h rest period (Light). Fat oxidation was measured using indirect calorimetry and plotted as a function of exercise intensity during Incr1 and Incr2. A sinusoidal equation, including 3 independent variables (dilatation, symmetry and translation), was used to characterize the fat oxidation kinetics and to determine the intensity (Fat(max)) that elicited the maximal fat oxidation (MFO) during Incr. After the Heavy and Light trials, Fat(max), MFO, and fat oxidation rates were significantly greater during Incr2 than Incr1 (p < 0.001). However, Δ (i.e., Incr2-Incr1) Fat(max), MFO, and fat oxidation rates were greater in the Heavy compared with the Light trial (p < 0.05). The fat oxidation kinetics during Incr2(Heavy) showed a greater dilatation and rightward asymmetry than Incr1(Heavy), whereas only a greater dilatation was observed in Incr2(Light) (p < 0.05). This study showed that although to a lesser extent in the Light trial, both prior exercise sessions led to an increase in Fat(max), MFO, and absolute fat oxidation rates during Incr2, inducing significant changes in the shape of the fat oxidation kinetics.

Interest rates mapping

The present study deals with the analysis and mapping of Swiss franc interest rates. Interest rates depend on time and maturity, defining term structure of the interest rate curves (IRC). In the present study IRC are considered in a two-dimensional feature space - time and maturity. Exploratory data analysis includes a variety of tools widely used in econophysics and geostatistics. Geostatistical models and machine learning algorithms (multilayer perceptron and Support Vector Machines) were applied to produce interest rate maps. IR maps can be used for the visualisation and pattern perception purposes, to develop and to explore economical hypotheses, to produce dynamic asset-liability simulations and for financial risk assessments. The feasibility of an application of interest rates mapping approach for the IRC forecasting is considered as well. (C) 2008 Elsevier B.V. All rights reserved.

A spatially explicit life cycle inventory of the global textile chain

Life cycle analyses (LCA) approaches require adaptation to reflect the increasing delocalization of production to emerging countries. This work addresses this challenge by establishing a country-level, spatially explicit life cycle inventory (LCI). This study comprises three separate dimensions. The first dimension is spatial: processes and emissions are allocated to the country in which they take place and modeled to take into account local factors. Emerging economies China and India are the location of production, the consumption occurs in Germany, an Organisation for Economic Cooperation and Development country. The second dimension is the product level: we consider two distinct textile garments, a cotton T-shirt and a polyester jacket, in order to highlight potential differences in the production and use phases. The third dimension is the inventory composition: we track CO2, SO2, NO (x), and particulates, four major atmospheric pollutants, as well as energy use. This third dimension enriches the analysis of the spatial differentiation (first dimension) and distinct products (second dimension). We describe the textile production and use processes and define a functional unit for a garment. We then model important processes using a hierarchy of preferential data sources. We place special emphasis on the modeling of the principal local energy processes: electricity and transport in emerging countries. The spatially explicit inventory is disaggregated by country of location of the emissions and analyzed according to the dimensions of the study: location, product, and pollutant. The inventory shows striking differences between the two products considered as well as between the different pollutants considered. For the T-shirt, over 70% of the energy use and CO2 emissions occur in the consuming country, whereas for the jacket, more than 70% occur in the producing country. This reversal of proportions is due to differences in the use phase of the garments. For SO2, in contrast, over two thirds of the emissions occur in the country of production for both T-shirt and jacket. The difference in emission patterns between CO2 and SO2 is due to local electricity processes, justifying our emphasis on local energy infrastructure. The complexity of considering differences in location, product, and pollutant is rewarded by a much richer understanding of a global production-consumption chain. The inclusion of two different products in the LCI highlights the importance of the definition of a product's functional unit in the analysis and implications of results. Several use-phase scenarios demonstrate the importance of consumer behavior over equipment efficiency. The spatial emission patterns of the different pollutants allow us to understand the role of various energy infrastructure elements. The emission patterns furthermore inform the debate on the Environmental Kuznets Curve, which applies only to pollutants which can be easily filtered and does not take into account the effects of production displacement. We also discuss the appropriateness and limitations of applying the LCA methodology in a global context, especially in developing countries. Our spatial LCI method yields important insights in the quantity and pattern of emissions due to different product life cycle stages, dependent on the local technology, emphasizing the importance of consumer behavior. From a life cycle perspective, consumer education promoting air-drying and cool washing is more important than efficient appliances. Spatial LCI with country-specific data is a promising method, necessary for the challenges of globalized production-consumption chains. We recommend inventory reporting of final energy forms, such as electricity, and modular LCA databases, which would allow the easy modification of underlying energy infrastructure.

Usages de l'eau à Gaya (Niger) : entre fortes potentialités et contraintes majeures

Résumé: Le département de Gaya, cadre de notre étude, est situé au sud-ouest de la république du Niger. Il dispose d'un important potentiel hydrique composé des eaux de surface (une centaine de mares permanentes, le fleuve Niger sur 106 km) et de sept aquifères superposés comprenant des nappes de subsurface (affleurantes par endroit) et des nappes artésiennes. L'étude sur les usages de l'eau à Gaya a été menée à travers plusieurs axes centrés sur l'estimation et la répartition spatiale des ressources en eau, le cadre juridique et institutionnel régulant leur mise en valeur, les différents secteurs d'utilisation de l'eau ainsi que les contraintes affectant cette utilisation. L'usage de la cartographie à travers les SIG dans le traitement et l'analyse des données, couplée à notre expérience d'une dizaine d'année de travaux dans la région, a permis de dresser des synthèses richement illustrées permettant de mieux comprendre tous les enjeux liés à la problématique des usages de l'eau dans cette partie du Niger. Contrairement à la vision que l'on a traditionnellement du Sahel où le manque d'eau constitue une des contraintes majeures au développement, ici des conditions locales particulières contredisent ce cliché et transposent le débat sur un autre plan. Il s'agit de la maîtrise de l'eau au niveau local à travers l'élaboration d'une politique appropriée qui tienne compte non seulement des spécificités locales de la ressource, mais aussi des différents types d'usages. La politique de l'eau au Niger, définie selon le Schéma directeur de mise en valeur et de gestion des ressources en eau, à travers la mise en place d'un important arsenal juridique et institutionnel, a eu le mérite de tracer un canevas sur la question, mais a montré ses limites au niveau pratique après dix ans d'essai. En effet au niveau de Gaya, ni l'Etat ni les partenaires au développement (bailleurs de fonds extérieurs) n'ont tenu compte des caractéristiques locales de la ressource ou du contexte socioéconomique particulier de la région. Ce qui a entraîné la réalisation d'infrastructures inadaptées aux réalités hydrogéologiques locales ainsi que des choix inappropriés au niveau de certains aménagements. En dépit de l'abondance de la ressource, son accès tant au niveau quantitatif que qualitatif, reste difficile pour une grande partie des acteurs ruraux. Les différents handicaps rencontrés dans la mise en valeur des ressources en eau résultent de cette incohérence de la politique nationale de l'eau, mais aussi de la difficulté de son application sur le terrain où persiste un pluralisme juridique caractérisé par la cohabitation de deux systèmes de régulation à savoir les droits coutumiers et la législation moderne. Ces différents éléments mis en évidence dans cette étude sur la zone de Gaya pourraient servir de base pour un meilleur aménagement des ressources en eau dans le cadre plus large d'une politique d'aménagement du territoire prenant en compte tous les facteurs tant physiques que socioéconomiques de la région. Abstract: The department of Gaya, in which this study was done, is located in the SW area of the Republic of Niger. It has an important hydrological potential composed of surface water (approximately 100 permanent ponds, 106 km of the Niger River) and 7 bodies of underground water sources including sub-surface and artresan wells. This study of the exploitation of wtaer in Gaya has been carried out employing several parameters based on: the estimation and spacial distribution of water ressources, the juridic and institutional rules governing their utilisation and the various constraints affecting this exploitation. The use of mapping when treating and analysing data, coupled with ten years personel field experience, resulted in a richly illustrated synthesis of this data. This, in turn, led to a better comprehension of all the factors related to problems of water utilisation in this particular region of Niger. Contrary to the generally accepted view that the lack of water ressources is a major contributing factor to the lack of development in the Sahel, in Gaya the local conditions contradict this statement. In this region, and at the local level, the proper use of water is based on the elaboration of an appropriate policy which takes into account not only the local specifics of water ressources but the various types of water utilsation as well. Local use of water and water ressources are dependant on established rules. Water policy in Niger is defined by the General Schema based on an important institutional and judicary arsenal of rules and regulations. However, after a ten-year trial period, this system was shown to have its limitations. In Gaya, neither the State nor the development agencies took into consideration local characteristics nor the socio-economic context of the region. This, in turn, resulted in putting in place infrastructures that were not adapted to local hydrogeological realities as well as inappropriate choices in land planning and development. In spite of the abundance of water ressources, access to them remains difficult for most of the rural population. The various difficulties encountered are the result of incoherent water policies on a national level as well as the lack of practical application in this area. This is due to a double judicary system where two regulatory systems co-exist:traditional laws and modern legislation. the different elements brought out by this study could serve as a basis for a better utilisation of water ressources on a larger scale in which land planning and development policies would take into consideration all the physcial as well as the socio-economical factors of this region.

Label-free detection of tobramycin in serum by transmission-localized surface plasmon resonance.

In order to improve the efficacy and safety of treatments, drug dosage needs to be adjusted to the actual needs of each patient in a truly personalized medicine approach. Key for widespread dosage adjustment is the availability of point-of-care devices able to measure plasma drug concentration in a simple, automated, and cost-effective fashion. In the present work, we introduce and test a portable, palm-sized transmission-localized surface plasmon resonance (T-LSPR) setup, comprised of off-the-shelf components and coupled with DNA-based aptamers specific to the antibiotic tobramycin (467 Da). The core of the T-LSPR setup are aptamer-functionalized gold nanoislands (NIs) deposited on a glass slide covered with fluorine-doped tin oxide (FTO), which acts as a biosensor. The gold NIs exhibit localized plasmon resonance in the visible range matching the sensitivity of the complementary metal oxide semiconductor (CMOS) image sensor employed as a light detector. The combination of gold NIs on the FTO substrate, causing NIs size and pattern irregularity, might reduce the overall sensitivity but confers extremely high stability in high-ionic solutions, allowing it to withstand numerous regeneration cycles without sensing losses. With this rather simple T-LSPR setup, we show real-time label-free detection of tobramycin in buffer, measuring concentrations down to 0.5 μM. We determined an affinity constant of the aptamer-tobramycin pair consistent with the value obtained using a commercial propagating-wave based SPR. Moreover, our label-free system can detect tobramycin in filtered undiluted blood serum, measuring concentrations down to 10 μM with a theoretical detection limit of 3.4 μM. While the association signal of tobramycin onto the aptamer is masked by the serum injection, the quantification of the captured tobramycin is possible during the dissociation phase and leads to a linear calibration curve for the concentrations over the tested range (10-80 μM). The plasmon shift following surface binding is calculated in terms of both plasmon peak location and hue, with the latter allowing faster data elaboration and real-time display of the results. The presented T-LSPR system shows for the first time label-free direct detection and quantification of a small molecule in the complex matrix of filtered undiluted blood serum. Its uncomplicated construction and compact size, together with the remarkable performances, represent a leap forward toward effective point-of-care devices for therapeutic drug concentration monitoring.