Lors du dimensionnement des ouvrages en terre : remblais, digues, on observe que la plupart des matériaux sont compactés à l’optimum Proctor ou coté humide. En général, ce compactage implique que le sol se trouve dans un état où le degré de saturation est très élevé. Cruz et al (1985) ont montré qu'à un degré de saturation élevé (supérieur à 85%, voire 90% dans le cas de certains sols), la phase liquide est continue alors que l’air présent sous forme de bulles est occlus ; ce qui rend le comportement du sol complexe. L’élaboration d’un modèle de comportement pour ce type de sols nécessite une compréhension approfondie des phénomènes physico-mécaniques intervenant au sein de l’air occlus, de l'eau liquide contenant de l'air dissous et du squelette solide. Dans ce sens, un nouveau modèle hydromécanique a été développé. Ce modèle prend en compte le comportement physico-mécanique et la cinématique propre de chacun des constituants du milieu polyphasé (eau liquide, air dissous, air sous forme gazeuse et matrice solide). En particulier, dans ce modèle, nous tenons compte de la tension de surface, de la migration des phases gazeuse et liquide qui ont des impacts importants sur le comportement mécanique des sols. Le développement du modèle conduit à un système d’équations aux dérivées partielles fortement non linéaire qui peut être résolu numériquement en utilisant la méthode des éléments finis. Ce nouveau modèle a été implémenté dans un code de calcul écrit en C++ « Hydromech », développé à l'origine par Pereira (2005), qui permet de simuler les essais oedométriques suivant différents trajets de chargement hydromécanique. En particulier, ce code de calcul permet de simuler de façon cohérente la transition entre différents régimes de saturation, aussi bien dans l'espace (translation progressive d'une frontière entre deux régimes voisins) que dans le temps (passage d'un régime à l'autre en un point donné) ; ce qui constitue un problème de modélisation difficile. Les études numériques réalisées montrent que ce modèle donne des résultats cohérents et mettent en évidence sa capacité à simuler avec précision le comportement hydromécanique des sols quasi-saturés comportant de l'air occlus. / The behaviour of quasi-saturated materials is an important factor to be considered when designing cuttings and embankments in which earthwork materials are compacted to the optimum proctor density. Typically, soil compaction is performed at the optimum Proctor or on the wet side of the optimum, which means that the soil is in a highly saturated state. Cruz et al (1985) have shown that at a high degree of saturation (greater than 85% or even 90% in the case of certain soils), the liquid phase is continuous whereas the gas phase in the form of entrapped air bubbles is discontinuous. It is the presence of the entrapped air bubbles which makes the soil behaviour complex. The construction of a theoretical model for this type of soils requires the consideration of various physical-mechanical phenomena and their couplings occurring within the tri-phasic medium consisting of the solid grains, liquid water containing dissolved air and the entrapped air bubbles. In this sense, a new hydromechanical model has been developed that takes into account the physical-mechanical interactions between different phases as well as the kinematics of each constituent (liquid water, dissolved air, gaseous air and solid grains). In particular, the model accounts for the interfacial tension, migration of gaseous and liquid phases, which have important impacts on the mechanical behaviour. The development leads to a system of highly non-linear partial differential equations which can be solved numerically using the finite element method. This new model has been implemented in a numerical code “Hydromech” written in C++, developed originally by Pereira (2005) that has been used to simulate oedometer tests with different hydromechanical loading paths. In particular, this code allows to simulate consistently the transition across different regimes of saturation, both with respect to space (progressive translation of a boundary between two neighbouring regimes) and to time (transition of one regime to another at a fixed material point); which constituted a difficult modelling problem at the start. Numerical studies carried out show that this model gives consistent results providing a clear demonstration of its ability to simulate with precision the hydro-mechanical behaviour of quasi-saturated soils containing entrapped air.

abstract: A dual-channel directional digital hearing aid (DHA) front-end using a fully differential difference amplifier (FDDA) based Microphone interface circuit (MIC) for a capacitive Micro Electro Mechanical Systems (MEMS) microphones and an adaptive-power analog font end (AFE) is presented. The Microphone interface circuit based on FDDA converts the capacitance variations into voltage signal, achieves a noise of 32 dB SPL (sound pressure level) and an SNR of 72 dB, additionally it also performs single to differential conversion allowing for fully differential analog signal chain. The analog front-end consists of 40dB VGA and a power scalable continuous time sigma delta ADC, with 68dB SNR dissipating 67u¬W from a 1.2V supply. The ADC implements a self calibrating feedback DAC, for calibrating the 2nd order non-linearity. The VGA and power scalable ADC is fabricated on 0.25 um CMOS TSMC process. The dual channels of the DHA are precisely matched and achieve about 0.5dB gain mismatch, resulting in greater than 5dB directivity index. This will enable a highly integrated and low power DHA / Dissertation/Thesis / Ph.D. Electrical Engineering 2011

Durant les 5 dernières décennies, les technologies CMOS se sont imposées comme méthode de fabrication principale pour les circuits semi-conducteurs intégrés avec notamment le transistor MOSFET. Néanmoins, la miniaturisation de ces transistors en technologie CMOS sur substrat massif atteint ses limites et a donc été arrêtée. Les filières FDSOI apparaissent comme une excellente alternative permettant une faible consommation et une excellente maîtrise des effets électrostatiques dans les transistors MOS, même pour les nœuds technologiques 14 et 28 nm. Cependant, la pente sous le seuil (60 mV/décade) du MOSFET ne peut pas être améliorée, ce qui limite la réduction de la tension d’alimentation. Cette restriction a motivé la recherche de composants innovants pouvant offrir des déclenchements abrupts tels que le Z2-FET (Zéro pente sous le seuil et Zéro ionisation par impact), Z2-FET DGP (avec double Ground Plane) et Z3-FET (Zéro grille avant). Grace à leurs caractéristiques intéressantes (déclenchement abrupte, faible courant de fuite, tension de déclenchement ajustable, rapport de courant ION/IOFF élevé), les dispositifs à modulation de bandes peuvent être utilisés dans différentes applications. Dans ce travail, nous nous sommes concentrés sur la protection contre les décharges électrostatiques (ESD), la mémoire DRAM embarquée sans capacité de stockage, et les interrupteurs logiques. L’étude des mécanismes statique et transitoire ainsi que des performances de ces composants a été réalisée grâce à des simulations TCAD détaillées, validées systématiquement par des résultats expérimentaux. Un modèle de potentiel de surface pour les trois dispositifs est également fourni. / During the past 5 decades, Complementary Metal Oxide Semiconductor (CMOS) technology was the dominant fabrication method for semiconductor integrated circuits where Metal Oxide Semiconductor Field Effect Transistor (MOSFET) was and still is the central component. Nonetheless, the continued physical downscaling of these transistors in CMOS bulk technology is suffering limitations and has been stopped nowadays. Fully Depleted Silicon-On-Insulator (FDSOI) technology appears as an excellent alternative that offers low-power consumption and improved electrostatic control for MOS transistors even in very advanced nodes (14 nm and 28 nm). However, the 60 mV/decade subthreshold slope of MOSFET is still unbreakable which limits the supply voltage reduction. This motivated us to explore alternative devices with sharp-switching: Z2-FET (Zero subthreshold slope and Zero impact ionization), Z2-FET DGP (with Dual Ground Planes) and Z3-FET (Zero front-gate). Thanks to their attractive characteristics (sharp switch, low leakage current, adjustable triggering voltage and high current ratio ION/IOFF), band-modulation devices are envisioned for multiple applications. In this work, we focused on Electro-Static Discharge (ESD) protection, capacitor-less Dynamic Random Access Memory and fast logic switch. The DC and transient operation mechanisms as well as the device performance are investigated in details with TCAD simulations and validated with systematic experimental results. A compact model of surface potential distribution for all Z-FET family devices is also given.

Thailand has been often affected by severe flood events over the past century. The 2011’s Thailand Flood Catastrophe was the costliest in country’s history, and it was ranked to be the second most damaging natural hazard in the world in terms of economic losses. The Chao Phraya River Basin was noted to be the most vulnerable area prone to flooding in Thailand. The dynamics of flood risk in the river basin have changed drastically over the past fifty years. In particular, flood exposure increased due to rapid urbanization and population growth. Since 2012, integrated flood risk management has been addressed to be the major framework of water-related disasters with the goal of losses and damage reductions. However, there is currently little research in Thailand on how to quantify flood risks and mitigate flood inundation damage on the relation between the occurrence of flood events and their consequential socio-economic implications. In this study, a tradition method in flood risk assessment is implemented by integrating 2D hydrodynamic modelling and the assessment of socio-economic impact of floods into the Chao Phraya River Basin. More specifically, the fully 2D version of the LISFLOOD-FP model code was used to model flood inundation processes. The output of the model was then used to map inundation depth and assess the levels of physical/environmental risk associated to flood hazards on multiple receptors/elements at risk. The European Flood Directive and the KULTURisk methodology were applied to quantify flood risks in monetary terms for residential, industrial, and agricultural sectors. The 2011 flood event was used for model calibration, while a hypothetical flood event with a return period of 100 years was simulated to identify the potential flood losses. Depth-damage functions comprising of JRC-ASIA, the Flemish, and JICA models were used to estimate potential damage for residential and industrial structures. The results showed that LISFLOOD-FP could satisfactorily reproduce the flood inundation extent obtained from satellite imagery in 2011. The model performance (Critical Success Index or F1) was of 56%, with a Bias of 112%. The latter meant the total inundated area was 12% larger than flood extent’s observation. Moreover, the model could simulate flood levels with overall Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) of 2.03 m a.s.l. and 1.78 m a.s.l., respectively. For the estimation of flood damage and losses, the Flemish model showed the strongest agreement with the reported flood damage in the residential sector, while JICA-ASIA model underestimated flood damage for industrial sector by just 1%. The KULTURisk methodology also well-estimated crop losses in the 2011 event which an overestimation about 21% from the reported value. Apart from that, fully 2D numerical method could not perfectly represent 1-in-100 year flood inundation due to non-consideration of important features such as the precise river channel topography, hydraulic infrastructures, and flood protection schemes in the river basin. Lack of such features results in an overestimation of flood damage and losses for 1-in-100 year flood comparing to the national flood hazard map and damage assessment which are simulated and estimated by JICA’s study. Such features can be better handled by using a coupled 1D/2D numerical method in order to simulate flood inundation extent more realistically and estimate flood losses. This could help the Thai government to better prepare a budget for flood risk prevention. In addition, even if the Flemish model indicates a good representation of relative flood damage to housing structures, the government should establish depth-damage curves specific for Thailand.

Volume measurements of timber loads is done in conjunction with timber trade. When dealing with goods of major economic values such as these, it is important to achieve an impartial and fair assessment when determining price-based volumes. With the help of Saab’s missile targeting technology, CIND AB develops products for digital volume measurement of timber loads. Currently there is a system in operation that automatically reconstructs timber trucks in motion to create measurable images of them. Future iterations of the system is expected to fully automate the scaling by generating a volumetric representation of the timber and calculate its external gross volume. The first challenge towards this development is to separate the timber load from the truck. This thesis aims to evaluate and implement appropriate method for semantic pixel-wise segmentation of timber loads in real time. Image segmentation is a classic but difficult problem in computer vision. To achieve greater robustness, it is therefore important to carefully study and make use of the conditions given by the existing system. Variations in timber type, truck type and packing together create unique combinations that the system must be able to handle. The system must work around the clock in different weather conditions while maintaining high precision and performance.

Este trabalho mostra a comparação dos efeitos das falhas provocadas pelos Single-Event Effects em dispositivos 28nm FDSOI, 28nm FDSOI High-K e 32nm Bulk CMOS e células de memória 6T SRAM feitas com estes dispositivos. Para conseguir isso, foram usadas ferramentas TCAD para simular falhas transientes devido a impacto de íons pesados a nível dispositivo e nível circuito. As simulações neste ambiente tem como vantagem a simulação dos fatos e mecanismos que produz as falhas transientes e seus efeitos nos dispositivos, além de também servir para projetar virtualmente estes dispositivos e caraterizar eles para estas simulações. Neste caso, foram projetados três dispositivos para simulação: um transistor NMOS de 32nm Bulk, um transistor NMOS de 28nm FDSOI e um transistor NMOS de 28nm FDSOI High-K para fazer comparações entre eles. Estes dispositivos foram projetados, caraterizados e testados contra o impacto de íons pesados a níveis dispositivo e circuito. Como resultado obtido, transistor Bulk de 32nm teve, no pior caso, uma carga coletada de 7.57 e 7.19 vezes maior que a carga coletada pelo dispositivo FDSOI de 28nm e FDSOI High-K de 28nm respectivamente atingido pelo mesmo íon pesado de 100MeV-cm2/mg. Com estes dados foi possível modelar o comportamento da carga coletada de ambos dispositivos usando este íon pesado, atingindo os terminais de Fonte e Dreno em distintos lugares e ângulos. Usando a mesma ferramenta e os dados obtidos de carga coletada pelos testes anteriores, foram projetadas células de memória SRAM de 6 transistores. Isso foi para testar elas contra os efeitos do impacto de íons pesados nos transistores NMOS de armazenagem da dados. Neste caso, a Transferência Linear de Energia (LET) do íon necessária para fazer que o dado armazenado na SRAM Bulk mude é 12.8 vezes maior que no caso da SRAM FDSOI e 10 maior no caso da SRAM FDSOI High-K, embora a quantidade de carga coletada necessária para que o dado mude em ambas células seja quase a mesma. Com estes dados foi possível modelar os efeitos dos íons pesados em ambos circuitos, descobrir a Carga Crítica destes e qual é o mínimo LET necessário para que o dado armazenado nestas SRAMs mude. / This work shows a comparison of faults due to Single-Event Effects in 28nm Fully Depleted SOI (FDSOI), 28nm FDSOI High-K and 32nm Bulk CMOS devices, and in 6T SRAM memory cells made with these devices. To provide this, was used TCAD tools to simulate transient faults due to heavy ion impacts on device and circuit levels. The simulations in that environment have the advantage to simulate the facts and mechanisms which produce the transient faults and this effects on the electronic devices, it also allow to simulate the virtual device fabrication and to characterize them. In this case, two devices were created for the simulations: a 32nm Bulk NMOS transistor and a 28nm FDSOI NMOS transistor for compare them. These devices were created, characterized and tested against heavy ion impacts at device and circuit levels. The results show that 32nm Bulk transistor has, in the worst case, a collected charge 7.57 and 7.19 times greater than the 28nm FDSOI and 28nm FDSOI High-K respectively collected charge with the same 100MeV-cm2/mg heavy ion. With these data it was possible to model the behavior of the collected charge in both devices with the same heavy-ion, reach the Source and Drain Terminal in different places and angles. Using the same tools and the obtained collected charge data of previous simulations, it was designed 6 transistors SRAM Memory Cells. That is done to test these circuits against the heavy ion effects on the data-storage NMOS transistor. In this case, the necessary Ion Linear Energy Transfer (LET) to flip the Bulk SRAM is 12.8 greater than the FDSOI SRAM and 10 times greater than the FDSOI High- K SRAM case, although the amount of charge to flip the cells is almost the same in both cases. With these data it was possible to model the heavy-ion effects in both circuits, discover the Critical Charge of them and the minimum LET to flips these SRAMs.

L’activation à basse température est prometteuse pour l’intégration 3D séquentielle où lebudget thermique du transistor supérieur est limité (<650 ºC) pour ne pas dégrader letransistor inférieur, mais aussi dans le cas d’une intégration planaire afin d’atteindre des EOTultra fines et de contrôler le travail de sortie de la grille sans recourir à une intégration de type« gate-last ». Dans ce travail, l’activation par recroissance en phase solide (SPER) a étéétudiée afin de réduire le budget thermique de l’activation des dopants.L’activation à basse température présente plusieurs inconvénients. Les travauxprécédents montrent que les fuites de jonctions sont plus importantes dans ces dispositifs.Ensuite, des fortes désactivations de dopants ont été observées. Troisièmement, la faiblediffusion des dopants rend difficile la connexion des jonctions source et drain avec le canal.Dans ce travail, il est montré que dans un transistor FDSOI, l’augmentation des fuites dejonctions et la désactivation du Bore peuvent être évités grâce à la présence de l’oxyde enterré.De plus les conditions d’implantation ont été optimisées et les transistors activés à650 ºC atteignent les performances des transistors de référence. / Low temperature (LT) process is gaining interest in the frame of 3D sequentialintegration where limited thermal budget (<650 ºC) is needed for top FET to preserve bottomFET from any degradation and also in the standard planar integration for achieving ultra-thinEOT and work function control with high-k metal gate without gate-last integration scheme.In this work, LT Solid Phase Epitaxial Regrowth (SPER) has been investigated for reducingthe most critical thermal budget which is dopant activation.From previous works, LT activated devices face several challenges: First, higher junctionleakage limits their application to high performance devices. Secondly, strong deactivation ofthe metastable activated dopants was observed with post anneals. Thirdly, the dopant weakdiffusion makes it difficult to connect the channel with S/D.In this work, it is shown that the use of FDSOI enables to overcome junction leakage andBoron deactivation issues thanks to the defect cutting off and sinking effect of buried oxide.As a consequence, dopant deactivation in FDSOI devices is no longer an issue. Finally,implants conditions of LT transistors have been optimized to reach similar performance thanits standard high temperature counterparts.

Este trabalho mostra a comparação dos efeitos das falhas provocadas pelos Single-Event Effects em dispositivos 28nm FDSOI, 28nm FDSOI High-K e 32nm Bulk CMOS e células de memória 6T SRAM feitas com estes dispositivos. Para conseguir isso, foram usadas ferramentas TCAD para simular falhas transientes devido a impacto de íons pesados a nível dispositivo e nível circuito. As simulações neste ambiente tem como vantagem a simulação dos fatos e mecanismos que produz as falhas transientes e seus efeitos nos dispositivos, além de também servir para projetar virtualmente estes dispositivos e caraterizar eles para estas simulações. Neste caso, foram projetados três dispositivos para simulação: um transistor NMOS de 32nm Bulk, um transistor NMOS de 28nm FDSOI e um transistor NMOS de 28nm FDSOI High-K para fazer comparações entre eles. Estes dispositivos foram projetados, caraterizados e testados contra o impacto de íons pesados a níveis dispositivo e circuito. Como resultado obtido, transistor Bulk de 32nm teve, no pior caso, uma carga coletada de 7.57 e 7.19 vezes maior que a carga coletada pelo dispositivo FDSOI de 28nm e FDSOI High-K de 28nm respectivamente atingido pelo mesmo íon pesado de 100MeV-cm2/mg. Com estes dados foi possível modelar o comportamento da carga coletada de ambos dispositivos usando este íon pesado, atingindo os terminais de Fonte e Dreno em distintos lugares e ângulos. Usando a mesma ferramenta e os dados obtidos de carga coletada pelos testes anteriores, foram projetadas células de memória SRAM de 6 transistores. Isso foi para testar elas contra os efeitos do impacto de íons pesados nos transistores NMOS de armazenagem da dados. Neste caso, a Transferência Linear de Energia (LET) do íon necessária para fazer que o dado armazenado na SRAM Bulk mude é 12.8 vezes maior que no caso da SRAM FDSOI e 10 maior no caso da SRAM FDSOI High-K, embora a quantidade de carga coletada necessária para que o dado mude em ambas células seja quase a mesma. Com estes dados foi possível modelar os efeitos dos íons pesados em ambos circuitos, descobrir a Carga Crítica destes e qual é o mínimo LET necessário para que o dado armazenado nestas SRAMs mude. / This work shows a comparison of faults due to Single-Event Effects in 28nm Fully Depleted SOI (FDSOI), 28nm FDSOI High-K and 32nm Bulk CMOS devices, and in 6T SRAM memory cells made with these devices. To provide this, was used TCAD tools to simulate transient faults due to heavy ion impacts on device and circuit levels. The simulations in that environment have the advantage to simulate the facts and mechanisms which produce the transient faults and this effects on the electronic devices, it also allow to simulate the virtual device fabrication and to characterize them. In this case, two devices were created for the simulations: a 32nm Bulk NMOS transistor and a 28nm FDSOI NMOS transistor for compare them. These devices were created, characterized and tested against heavy ion impacts at device and circuit levels. The results show that 32nm Bulk transistor has, in the worst case, a collected charge 7.57 and 7.19 times greater than the 28nm FDSOI and 28nm FDSOI High-K respectively collected charge with the same 100MeV-cm2/mg heavy ion. With these data it was possible to model the behavior of the collected charge in both devices with the same heavy-ion, reach the Source and Drain Terminal in different places and angles. Using the same tools and the obtained collected charge data of previous simulations, it was designed 6 transistors SRAM Memory Cells. That is done to test these circuits against the heavy ion effects on the data-storage NMOS transistor. In this case, the necessary Ion Linear Energy Transfer (LET) to flip the Bulk SRAM is 12.8 greater than the FDSOI SRAM and 10 times greater than the FDSOI High- K SRAM case, although the amount of charge to flip the cells is almost the same in both cases. With these data it was possible to model the heavy-ion effects in both circuits, discover the Critical Charge of them and the minimum LET to flips these SRAMs.

Esta tese se concentra em aplicar os modelos de mistura de distribuições na segmentação de imagens SAR polarimétricas multi-look. Dentro deste contexto, utilizou-se o algoritmo SEM em conjunto com os estimadores obtidos pelo método dos momentos para calcular as estimativas dos parâmetros do modelo de mistura das distribuições Wishart, Kp ou G0p. Cada uma destas distribuições possui parâmetros específicos que as diferem no ajuste dos dados com graus de homogeneidade variados. A distribuição Wishart descreve bem regiões com características mais homogêneas, como cultivo. Esta distribuição é muito utilizada na análise de dados SAR polarimétricos multi-look. As distribuições Kp e G0p possuem um parâmetro de rugosidade que as permitem descrever tanto regiões mais heterogêneas, como vegetação e áreas urbanas, quanto regiões homogêneas. Além dos modelos de mistura de uma única família de distribuições, também foi analisado o caso de um dicionário contendo as três famílias. Há comparações do método SEM proposto para os diferentes modelos com os métodos da literatura k-médias e EM utilizando imagens reais da banda L. O método SEM com a mistura de distribuições G0p forneceu os melhores resultados quando os outliers da imagem são desconsiderados. A distribuição G0p foi a mais flexível ao ajuste dos diferentes tipos de alvo. A distribuição Wishart foi robusta às diferentes inicializações. O método k-médias com a distribuição Wishart é robusto à segmentação de imagens contendo outliers, mas não é muito flexível à variabilidade das regiões heterogêneas. O modelo de mistura do dicionário de famílias melhora a log-verossimilhança do método SEM, mas apresenta resultados parecidos com os do modelo de mistura G0p. Para todos os tipos de inicialização e grupos, a distribuição G0p predominou no processo de seleção das distribuições do dicionário de famílias. / The main focus of this thesis consists of the application of mixture models in multi-look polarimetric SAR image segmentation. Within this context, the SEM algorithm, together with the method of moments, were applied in the estimation of the Wishart, Kp and G0p mixture model parameters. Each one of these distributions has specific parameters that allows fitting data with different degrees of homogeneity. The Wishart distribution is suitable for modeling homogeneous regions, like crop fields for example. This distribution is widely used in multi-look polarimetric SAR data analysis. The distributions Kp and G0p have a roughness parameter that allows them to describe both heterogeneous regions, as vegetation and urban areas, and homogeneous regions. Besides adopting mixture models of a single family of distributions, the use of a dictionary with all the three family of distributions was proposed and analyzed. Also, a comparison between the performance of the proposed SEM method, considering the different models in real L-band images and two widely known techniques described in literature (k-means and EM algorithms), are shown and discussed. The proposed SEM method, considering a G0p mixture model combined with a outlier removal stage, provided the best classication results. The G0p distribution was the most flexible for fitting the different kinds of data. The Wishart distribution was robust for different initializations. The k-means algorithm with Wishart distribution is robust for segmentation of SAR images containing outliers, but it is not so flexible to variabilities in heterogeneous regions. The mixture model considering the dictionary of distributions improves the SEM method log-likelihood, but presents similar results to those of G0p mixture model. For all types of initializations and clusters, the G0p prevailed in the distribution selection process of the dictionary of distributions.

This study focuses on and tries to identify areas where machine translation can be useful by examining translated fictional and non-fictional texts, and the extent to which these different text types are better or worse suited for machine translation.  It additionally evaluates the performance of the free online translation tool Google Translate (GT). The BLEU automatic evaluation metric for machine translation was used for this study, giving a score of 27.75 BLEU value for fictional texts and 32.16 for the non-fictional texts. The non-fictional texts are samples of law documents, (commercial) company reports, social science texts (religion, welfare, astronomy) and medicine. These texts were selected because of their degree of difficulty. The non-fictional sentences are longer than those of the fictional texts and in this regard MT systems have struggled. In spite of having longer sentences, the non-fictional texts got a higher BLUE score than the fictional ones. It is speculated that one reason for the higher score of non-fictional texts might be that more specific terminology is used in these texts, leaving less room for subjective interpretation than for the fictional texts. There are other levels of meaning at work in the fictional texts that the human translator needs to capture.