From Word Embeddings to Large Vocabulary Neural Machine Translation

Dans ce mémoire, nous examinons certaines propriétés des représentations distribuées de mots et nous proposons une technique pour élargir le vocabulaire des systèmes de traduction automatique neurale. En premier lieu, nous considérons un problème de résolution d'analogies bien connu et examinons l'effet de poids adaptés à la position, le choix de la fonction de combinaison et l'impact de l'apprentissage supervisé. Nous enchaînons en montrant que des représentations distribuées simples basées sur la traduction peuvent atteindre ou dépasser l'état de l'art sur le test de détection de synonymes TOEFL et sur le récent étalon-or SimLex-999. Finalament, motivé par d'impressionnants résultats obtenus avec des représentations distribuées issues de systèmes de traduction neurale à petit vocabulaire (30 000 mots), nous présentons une approche compatible à l'utilisation de cartes graphiques pour augmenter la taille du vocabulaire par plus d'un ordre de magnitude. Bien qu'originalement développée seulement pour obtenir les représentations distribuées, nous montrons que cette technique fonctionne plutôt bien sur des tâches de traduction, en particulier de l'anglais vers le français (WMT'14).

Assessment of PEG on Polymeric Particles Surface, a Key Step in Drug Carrier Translation

Injectable drug nanocarriers have greatly benefited in their clinical development from the addition of a superficial hydrophilic corona to improve their cargo pharmacokinetics. The most studied and used polymer for this purpose is poly(ethylene glycol), PEG. However, in spite of its wide use for over two decades now, there is no general consensus on the optimum PEG chain coverage-density and size required to escape from the mononuclear phagocyte system and to extend the circulation time. Moreover, cellular uptake and active targeting may have conflicting requirements in terms of surface properties of the nanocarriers which complicates even more the optimization process. These persistent issues can be largely attributed to the lack of straightforward characterization techniques to assess the coverage-density, the conformation or the thickness of a PEG layer grafted or adsorbed on a particulate drug carrier and is certainly one of the main reasons why so few clinical applications involving PEG coated particle-based drug delivery systems are under clinical trial so far. The objective of this review is to provide the reader with a brief description of the most relevant techniques used to assess qualitatively or quantitatively PEG chain coverage-density, conformation and layer thickness on polymeric nanoparticles. Emphasis has been made on polymeric particle (solid core) either made of copolymers containing PEG chains or modified after particle formation. Advantages and limitations of each technique are presented as well as methods to calculate PEG coverage-density and to investigate PEG chains conformation on the NP surface.

Frequent translation strategies used by Iranian translators in subtitles to translate metaphors

En tant que moyen de communication qui détermine les structures de base des langues, la métaphore se classe parmi les enjeux les plus importants du domaine de la traductologie, d’où la complexité de la rendre traduisible. La nature indirecte de la métaphore étant problématique, celle-ci exige une étude approfondie, surtout vis-à-vis le transfert de la culture d’une langue à une autre. Un survol de la littérature académique traitant des théories et des approches différentes en traductologie révèle une façon de rendre possible la traduction de la métaphore. Cette étude traite de la traduction des métaphores du perse aux sous-titres anglais. Elle indique également les approches et les stratégies appliquées à la traduction des expressions linguistiques métaphoriques telles que proposées par Newmark (1988). En se fondant sur des exemples authentiques tirés des textes sources (deux films perses bien connus) et leurs traductions anglaises, cette étude démontre la façon dont certains traducteurs iraniens traduisent des expressions métaphoriques en se servant d’une des sept stratégies proposées par Newmark. Elle explique également comment certains éléments (par ex. les difficultés de la traduction) ont été influents sur leurs conclusions, tout en soutenant l’hypothèse que la reproduction d’une même image d’une langue source dans une langue cible ainsi que l’effacement sont respectivement la plus fréquente et la moins fréquente des stratégies employées par les sous-titreurs iraniens.

A STUDY ON SOFTWARE DEVELOPMENT PROJECT RISK, RISK MANAGEMENT, PROJECT OUTCOMES AND THEIR INTER-RELATIONSHIP

This research was undertaken with an objective of studying software development project risk, risk management, project outcomes and their inter-relationship in the Indian context. Validated instruments were used to measure risk, risk management and project outcome in software development projects undertaken in India. A second order factor model was developed for risk with five first order factors. Risk management was also identified as a second order construct with four first order factors. These structures were validated using confirmatory factor analysis. Variation in risk across categories of select organization / project characteristics was studied through a series of one way ANOVA tests. Regression model was developed for each of the risk factors by linking it to risk management factors and project /organization characteristics. Similarly regression models were developed for the project outcome measures linking them to risk factors. Integrated models linking risk factors, risk management factors and project outcome measures were tested through structural equation modeling. Quality of the software developed was seen to have a positive relationship with risk management and negative relationship with risk. The other outcome variables, namely time overrun and cost over run, had strong positive relationship with risk. Risk management did not have direct effect on overrun variables. Risk was seen to be acting as an intervening variable between risk management and overrun variables.

An Inter-digital Capacitive Electrode Modified as a Pressure Sensor

Inter-digital capacitive electrodes working as electric field sensors have been developed for touch panel applications. Evaluation circuits to convert variations in electric fields in such sensors into computer compatible data are commercially available. We report development of an Interdigital capacitive electrode working as a sensitive pressure sensor in the range 0-120 kPa. Essentially it is a touch/proximity sensor converted into a pressure sensor with a suitable elastomer buffer medium acting as the pressure transmitter. The performance of the sensor has been evaluated and reported. Such sensors can be made very economical in comparison to existing pressure sensors. Moreover, they are very convenient to be fabricated into sensor arrays involving a number of sensors for distributed pressure sensing applications such as in biomedical systems.

Inter- State Variations in Productivity trends and Technological Change in Indian industry

The present study entitled ‘Inter-State Variations in Manufacturing Productivity and Technological Changes in India’ covers a period of 38 years from l960 tol998-99. The study is mainly based on ASI data. The study starts with a discussion of the major facilitating factors of industrialization, namely, historical forces, public policy and infrastructure facilities. These are discussed in greater details in the context of our discussion on Perrox’s (1998) ‘growth pole’ and ‘development pole’, Hirschman’s (1958) ‘industrial centers’ and Myrdal’s ‘spread effect’ Most of the existing literature more or less agrees that the process of industrialization has not been unifonn in all Indian states. There has been a decline in inter-state industrial disparities over time. This aspect is dealt at some length in the third chapter. An important element that deserves detailed attention is the intra-regional differences in industrialisation. Regional industrialisation implies the emergence of a few focal points and industrial regions. Calcutta, Bombay and Madras were the initial focal points. Later other centers like Bangalore, Amritsar, Ahemedabad etc. emerged as nodal points in other states. All major states account for focal points. The analysis made in the third chapter shows that industrial activities generally converge to one or two focal points and industrial regions have emerged out of the focal points in almost all states. One of the general features of these complexes and regions is that they approximately accommodate 50 to 75 percent of the total industrial units and workers in the state. Such convergence is seen hands in glow with urbanization. It was further seen that intra-regional industrial disparity comes down in industrial states like Maharashtra, Gujarat and Uttar Pradesh.

“Income, Livelihood and Education of Tribal Communities in Kerala – Exploring Inter Community Disparities

The overall objective of the study is to examine whether the tribal communities in Kerala can be considered a coherent group in terms of select indicators of development by focusing on nine major tribal communities. The study also aims to bring out the intercommunity differences if any in aspects of livelihood options and education level of the tribal communities in Kerala

Studies on a syntax based approach for translation between structurally different languages and the development of a prototype for Malayalam to English translation

This thesis summarizes the results on the studies on a syntax based approach for translation between Malayalam, one of Dravidian languages and English and also on the development of the major modules in building a prototype machine translation system from Malayalam to English. The development of the system is a pioneering effort in Malayalam language unattempted by previous researchers. The computational models chosen for the system is first of its kind for Malayalam language. An in depth study has been carried out in the design of the computational models and data structures needed for different modules: morphological analyzer , a parser, a syntactic structure transfer module and target language sentence generator required for the prototype system. The generation of list of part of speech tags, chunk tags and the hierarchical dependencies among the chunks required for the translation process also has been done. In the development process, the major goals are: (a) accuracy of translation (b) speed and (c) space. Accuracy-wise, smart tools for handling transfer grammar and translation standards including equivalent words, expressions, phrases and styles in the target language are to be developed. The grammar should be optimized with a view to obtaining a single correct parse and hence a single translated output. Speed-wise, innovative use of corpus analysis, efficient parsing algorithm, design of efficient Data Structure and run-time frequency-based rearrangement of the grammar which substantially reduces the parsing and generation time are required. The space requirement also has to be minimised

Syntactic Based Machine Translation from English to Malayalam

Due to the emergence of multiple language support on the Internet, machine translation (MT) technologies are indispensable to the communication between speakers using different languages. Recent research works have started to explore tree-based machine translation systems with syntactical and morphological information. This work aims the development of Syntactic Based Machine Translation from English to Malayalam by adding different case information during translation. The system identifies general rules for various sentence patterns in English. These rules are generated using the Parts Of Speech (POS) tag information of the texts. Word Reordering based on the Syntax Tree is used to improve the translation quality of the system. The system used Bilingual English –Malayalam dictionary for translation.

English to Malayalam Translation: A Statistical Approach

This paper underlines a methodology for translating text from English into the Dravidian language, Malayalam using statistical models. By using a monolingual Malayalam corpus and a bilingual English/Malayalam corpus in the training phase, the machine automatically generates Malayalam translations of English sentences. This paper also discusses a technique to improve the alignment model by incorporating the parts of speech information into the bilingual corpus. Removing the insignificant alignments from the sentence pairs by this approach has ensured better training results. Pre-processing techniques like suffix separation from the Malayalam corpus and stop word elimination from the bilingual corpus also proved to be effective in training. Various handcrafted rules designed for the suffix separation process which can be used as a guideline in implementing suffix separation in Malayalam language are also presented in this paper. The structural difference between the English Malayalam pair is resolved in the decoder by applying the order conversion rules. Experiments conducted on a sample corpus have generated reasonably good Malayalam translations and the results are verified with F measure, BLEU and WER evaluation metrics

Handling OOV Words in Phrase-Based Statistical Machine Translation for Malayalam

Statistical Machine Translation (SMT) is one of the potential applications in the field of Natural Language Processing. The translation process in SMT is carried out by acquiring translation rules automatically from the parallel corpora. However, for many language pairs (e.g. Malayalam- English), they are available only in very limited quantities. Therefore, for these language pairs a huge portion of phrases encountered at run-time will be unknown. This paper focuses on methods for handling such out-of-vocabulary (OOV) words in Malayalam that cannot be translated to English using conventional phrase-based statistical machine translation systems. The OOV words in the source sentence are pre-processed to obtain the root word and its suffix. Different inflected forms of the OOV root are generated and a match is looked up for the word variants in the phrase translation table of the translation model. A Vocabulary filter is used to choose the best among the translations of these word variants by finding the unigram count. A match for the OOV suffix is also looked up in the phrase entries and the target translations are filtered out. Structuring of the filtered phrases is done and SMT translation model is extended by adding OOV with its new phrase translations. By the results of the manual evaluation done it is observed that amount of OOV words in the input has been reduced considerably

Techniques to Improve the word alignments in Statistical Machine Translation from English to Malayalam

In Statistical Machine Translation from English to Malayalam, an unseen English sentence is translated into its equivalent Malayalam translation using statistical models like translation model, language model and a decoder. A parallel corpus of English-Malayalam is used in the training phase. Word to word alignments has to be set up among the sentence pairs of the source and target language before subjecting them for training. This paper is deals with the techniques which can be adopted for improving the alignment model of SMT. Incorporating the parts of speech information into the bilingual corpus has eliminated many of the insignificant alignments. Also identifying the name entities and cognates present in the sentence pairs has proved to be advantageous while setting up the alignments. Moreover, reduction of the unwanted alignments has brought in better training results. Experiments conducted on a sample corpus have generated reasonably good Malayalam translations and the results are verified with F measure, BLEU and WER evaluation metrics