Cross-Lingual spoken language understanding from unaligned data using discriminative classification models and machine translation

This paper investigates several approaches to bootstrapping a new spoken language understanding (SLU) component in a target language given a large dataset of semantically-annotated utterances in some other source language. The aim is to reduce the cost associated with porting a spoken dialogue system from one language to another by minimising the amount of data required in the target language. Since word-level semantic annotations are costly, Semantic Tuple Classifiers (STCs) are used in conjunction with statistical machine translation models both of which are trained from unaligned data to further reduce development time. The paper presents experiments in which a French SLU component in the tourist information domain is bootstrapped from English data. Results show that training STCs on automatically translated data produced the best performance for predicting the utterance's dialogue act type, however individual slot/value pairs are best predicted by training STCs on the source language and using them to decode translated utterances. © 2010 ISCA.

Unsupervised intra-lingual and cross-lingual speaker adaptation for HMM-based speech synthesis using two-pass decision tree construction

Hidden Markov model (HMM)-based speech synthesis systems possess several advantages over concatenative synthesis systems. One such advantage is the relative ease with which HMM-based systems are adapted to speakers not present in the training dataset. Speaker adaptation methods used in the field of HMM-based automatic speech recognition (ASR) are adopted for this task. In the case of unsupervised speaker adaptation, previous work has used a supplementary set of acoustic models to estimate the transcription of the adaptation data. This paper first presents an approach to the unsupervised speaker adaptation task for HMM-based speech synthesis models which avoids the need for such supplementary acoustic models. This is achieved by defining a mapping between HMM-based synthesis models and ASR-style models, via a two-pass decision tree construction process. Second, it is shown that this mapping also enables unsupervised adaptation of HMM-based speech synthesis models without the need to perform linguistic analysis of the estimated transcription of the adaptation data. Third, this paper demonstrates how this technique lends itself to the task of unsupervised cross-lingual adaptation of HMM-based speech synthesis models, and explains the advantages of such an approach. Finally, listener evaluations reveal that the proposed unsupervised adaptation methods deliver performance approaching that of supervised adaptation.

A smart load appliance application using a single compressor fridge-freezer to support grid operations

This paper describes a fridge-freezer smart load model, which responds to external signals from the wholesale electricity market to support grid operations while switching the fridge-freezer on and off to maintain optimum operations for the owner. The key parameters of the model are the appliance dimensions, thermal mass, the fridge and freezer thermal time constants and the compressor power consumption. The model demonstrates that control strategies help to minimise load at times when the grid is under stress from high demand, and shift some load to a lower wholesale price or when there is excess renewable power. Three control strategies are proposed, based on peak shaving and valley filling, price signals and wind availability.

Thermal performance in a conventional fireplace and in an inset appliance

The objective of this study is to compare the performance of the equipment currently employed in the domestic heating with firewood, the conventional fireplace and the inset appliance with close firedoors. For such, it was followed the European Standard EN13229:2001/A2:2004. Efficiency and heat output is determined and the major heat losses that penalize the appliance performance are identified and calculated. Tests in laboratory were developed in two inset appliances, and tests in situ in one conventional fireplace. One of the appliances uses only staging as primary air, and the other only grate air In inset appliances, with a heat output near 10 kW, the average efficiency varies between 67% and 73%, while in a conventional fireplace that value lies at 30%. In all these devices the major losses take place as sensible heat in the flue gases, 23% to 30 % in inset appliances and above 50% in a conventional fireplace. The second most important heat loss happens by chemical losses in the flue gases. It takes values near 17% in a conventional fireplace and may be disregarded in an inset appliance.

Discrepancia existente en el tamaño mesiodistal y vestíbulo-lingual del primer premolar y segundo premolar superior.

Tesis (Maestría en Ciencias Odontológicas con Especialidad en Ortodoncia) UANL, 2010.

Étude clinique randomisée prospective du taux de survie d'un fil lingual mandibulaire de rétention utilisant les méthodes de collage direct et indirect à court et moyen termes

Introduction : Après un traitement orthodontique, la rétention (ou contention) est essentielle pour éviter les récidives vers la malocclusion initiale. Le fil de rétention lingual est un appareil fixe, relativement facile à installer et bien accepté par les patients pour maintenir la position finale des dents antérieures inférieures. Étant de plus en plus utilisé, il devient important de s’assurer de sa fiabilité pour la stabilité de l’alignement dentaire. Objectif : Le but de cette étude clinique randomisée prospective est de déterminer le taux de survie d’un fil lingual mandibulaire de rétention en comparant les méthodes de collage direct et de collage indirect à court et moyen termes. Méthodologie : L’échantillon est constitué de 117 patients consécutifs aléatoirement distribués dans 2 groupes : collage direct (n=58) et collage indirect (n=59). Les fils torsadés de diamètre 0,0175’’ sont préformés par un technicien de laboratoire soit selon la méthode de collage direct, soit selon la méthode de collage indirect. Une matrice de transfert en silicone assure le positionnement précis du fil lingual en bouche. Assure® et Filtek™ Flow ont été utilisés pour le collage direct. Filtek™ Flow, Assure®, and Sondhi™ ont été utilisés pour le collage indirect. Les fils de rétention ont été évalués pour le décollement, l’infiltration, la distorsion et le bris à 2 mois (T1) et 6 mois (T2). Résultats : À T1, le taux de survie du fil de rétention est de 90,2% pour le groupe de collage direct, comparativement à 79,5% pour le groupe de collage indirect (p=0,232). À T2, le fil est resté intact pour 74,1% des participants dans le groupe de collage direct et pour 70,0% des participants dans le groupe de collage indirect (p=0,481). Les différences ne sont pas statistiquement significatives entre les 2 groupes. La fréquence du décollement est plus haute que les autres problèmes enregistrés à T1 (p<0,022), représentant 85,7% des échecs. À T2, le décollement est plus fréquent que la distorsion ou le bris (p<0,04), mais pas statistiquement plus fréquent que l’infiltration (p=0,109). Il représente alors 86,4% des échecs. Conclusion : Le décollement est la principale cause d’échec d’un fil de rétention lingual. Il n’y a pas de différence statistiquement significative du taux de survie d’un fil lingual mandibulaire de rétention entre les techniques de collage direct et de collage indirect à court et moyen termes.

English-Malayalam Cross-Lingual Information Retrieval – an Experience

This paper describes about an English-Malayalam Cross-Lingual Information Retrieval system. The system retrieves Malayalam documents in response to query given in English or Malayalam. Thus monolingual information retrieval is also supported in this system. Malayalam is one of the most prominent regional languages of Indian subcontinent. It is spoken by more than 37 million people and is the native language of Kerala state in India. Since we neither had any full-fledged online bilingual dictionary nor any parallel corpora to build the statistical lexicon, we used a bilingual dictionary developed in house for translation. Other language specific resources like Malayalam stemmer, Malayalam morphological root analyzer etc developed in house were used in this work