A dissertation submitted in fulfilment of the requirements for the award of

This paper presents a successful technique for creating a new language model (LM) that adapts the original target LM used by a machine translation (MT) system. This technique is especially useful for situations where there are very scarce resources for training the target side (Spanish Sign Language (LSE) in our case) in order to properly estimate the target LM, the Sign Language Model (SLM), used by the MT system. The technique uses information from the source language, Spanish in our task, and from the phrase-based translation matrix in order to create a new LM, estimated using web frequencies, which adapts the counts of the SLM through the Maximum A Posteriori method (MAP). The corpus consists of common used sentences spoken by an officer when assisting people in applying for, or renewing, the National Identification Document. The proposed technique allows relative reductions of 15.5 % on perplexity and 2.7% on WER for translation, which are close to half the maximum performance obtainable when only the LM is optimized. Index Terms: language model adaptation, machine translation, sign language, web counts.

In this paper we will summarize the work done in the PhD thesis that follows the same title as this document. In the thesis we propose different innovative, dynamic, and intelligent acceleration strategies applied to a development platform for reducing the design time of multimodal and multilingual dialogue systems and to improve the runtime modules. Throughout the paper we will describe the three different kinds of accelerations proposed, which are innovative with respect to current commercial and research platforms. The first kind of strategies was applied to the design platform in order to allow the prediction of the information required to complete the different aspects of the service. These strategies are mainly based on using the data model structure and database contents, as well as cumulative information obtained from the previous and sequential steps in the design. Thanks to them, the design is reduced, most of the times, to simple confirmations from the designer to the “proposals ” that the platform automatically provides. The second kind of strategies is the incorporation of a new adaptation algorithm to the language models used by a machine translation system that automatically translates system’s prompts (in audio or text) into an animated sequence in Sign Language for providing the designed service to deaf users using an avatar. Finally, the third kind is an innovative LID technique based on using a discriminative ranking of n-grams that allows the incorporation of contextual longer-span information into the language models used to identify the system needs to use to interact with the users of the service.