We present Minimum Bayes-Risk (MBR) decoding for statistical machine translation. This statistical approach aims to minimize expected loss of translation errors under loss functions that measure translation performance. We describe a hierarchy of loss functions that incorporate different levels of linguistic information from word strings, word-to-word alignments from an MT system, and syntactic structure from parse-trees of source and target language sentences. We report the performance of the MBR decoders on a Chinese-to-English translation task. Our results show that MBR decoding can be used to tune statistical MT performance for specific loss functions. 1

The goal of interactive machine translation is to improve the productivity of human translators. An interactive machine translation system operates as follows: the automatic system proposes a translation.

Our previous work on statistical language modeling introduced the use of probabilistic feedforward neural networks to help dealing with the curse of dimensionality. Training this model by maximum likelihood however requires for each example to perform as many network passes as there are words in the vocabulary. Inspired by the contrastive divergence model, we propose and evaluate sampling-based methods which require network passes only for the observed "positive example" and a few sampled negative example words. A very significant speed-up is obtained with an adaptive importance sampling.

Text prediction is a novel form of interactive machine translation that is well suited to skilled translators. It has the potential to assist in several ways: speeding typing, suggesting possible translations, and averting translator errors. However, recent evaluations of a prototype prediction system showed that predictions can also distract and hinder translators if made indiscriminately. We demonstrate an experimental prototype intended to address this problem by selecting the prediction that has maximal expected benefit to the user in any given context. This leads it to make longer predictions where it is more certain and shorter ones---or none at all---in contexts where it is less certain.

Abstract. We consider the problem of predicting how a user will continue a given initial text fragment. Intuitively, our goal is to develop a “tab-complete ” function for natural language, based on a model that is learned from text data. We consider two learning mechanisms that generate predictive models from collections of application-specific document collections: we develop an N-gram based completion method and discuss the application of instance-based learning. After developing evaluation metrics for this task, we empirically compare the model-based to the instance-based method and assess the predictability of call-center emails, personal emails, and weather reports. 1

We present a method for the computation of prefix probabilities for synchronous context-free grammars. Our framework is fairly general and relies on the combination of a simple, novel grammar transformation and standard techniques to bring grammars into normal forms.

phrase-based models for interactive computer-assisted translation

User preferences for automated assistance often vary widely, depending on the situation, and quality or presentation of help. Developing effective models to learn individual preferences online requires domain models that associate observations of user behavior with their utility functions, which in turn can be constructed using utility elicitation techniques. However, most elicitation methods ask for users ’ predicted utilities based on hypothetical scenarios rather than more realistic experienced utilities. This is especially true in interface customization, where users are asked to assess novel interface designs. We propose experiential utility elicitation methods for customization and compare these to predictive methods. As experienced utilities have been argued to better reflect true preferences in behavioral decision making, the purpose here is to investigate accurate and efficient procedures that are suitable for software domains. Unlike conventional elicitation, our results indicate that an experiential approach helps people understand stochastic outcomes, as well as better appreciate the sequential utility of intelligent assistance. 1

Abstract — We investigate factors related to interfacing a speech-to-speech translation device with multimodal capabilities. We evaluate the efficacy of the interactions using a measure for meaning transfer, we call concept score. We show that employing a multimodal interface improves translation quality, in this study, by 24%. We also show that while some users require perfect representation of what they said in order to allow transfer, others accept concept degradation to some extent, in median up to 20% in our experiments. An appropriate system strategy is required to recognize this behavior and guide users towards optimum performance points. For example, we show that appropriate feedback is required to guide the users in their choices of translation method, as 13 % of the choices users made are worse than the alternatives the system provided. I.

Given a Pattern Recognition task, Computer Assisted Pattern Recognition can be viewed as a series of solution proposals made by a computer system, followed by corrections made by a user, until an acceptable solution is found. For this kind of systems, the appropriate measure of performance is the expected number of corrections the user has to make. In the present work we study the special case when the solution proposals have a sequential nature. Some examples of this type of tasks are: language translation, speech transcription and handwriting text transcription. In all these cases the output (the solution proposal) is a sequence of symbols. In this framework it is assumed that the user corrects always the first error found in the proposed solution. As a consequence, the prefix of the proposed solution before the last error correction can be assumed error free in the next iteration. Nowadays, all the techniques in the literature relies in proposing, at each step, the most probable suffix given that a prefix of the “correct ” output is already known. Usually the computation of the conditional most probable output is an NP-Hard or an undecidable problem (and then we have to apply some approximations) or, in some simple cases, complex dynamic programming techniques should be used (usualy some variant of the Viterbi algorithm). In the present work we show that this strategy is not optimum when we are interested in minimizing the number of human interactions. Moreover we describe the optimum strategy that is simpler (and usually faster) to compute.