Vicente Alabau

Statistical Post-Editing of a Rule-Based Machine Translation System

Automatic post-editing (APE) systems aim at correcting the output of machine translation systems to produce better quality translations, i.e. produce translations can be manually post-edited with an increase in productivity. In this work, we present an APE system that uses statistical models to enhance a commercial rule-based machine translation (RBMT) system. In addition, a procedure for effortless human evaluation has been established. We have tested the APE system with two corpora of different complexity. For the Parliament corpus, we show that the APE system significantly complements and improves the RBMT system. Results for the Protocols corpus, although less conclusive, are promising as well. Finally, several possible sources of errors have been identified which will help develop future system enhancements.

Interactive machine translation using a web-based architecture

In this paper we present a new way of translating documents by using a Web-based system. An interactive approach is proposed as an alternative to post-editing the output of a machine translation system. In this approach, the users feedback is used to validate or to correct parts of the system output that allow the generation of improved versions of the rest of the output.

Confidence Measures for Error Correction in Interactive Transcription Handwritten Text

An effective approach to transcribe old text documents is to follow an interactive-predictive paradigm in which both, the system is guided by the human supervisor, and the supervisor is assisted by the system to complete the transcription task as efficiently as possible. In this paper, we focus on a particular system prototype called GIDOC, which can be seen as a first attempt to provide user-friendly, integrated support for interactive-predictive page layout analysis, text line detection and handwritten text transcription. More specifically, we focus on the handwriting recognition part of GIDOC, for which we propose the use of confidence measures to guide the human supervisor in locating possible system errors and deciding how to proceed. Empirical results are reported on two datasets showing that a word error rate not larger than a 10% can be achieved by only checking the 32% of words that are recognised with less confidence.

A multimodal predictive-interactive application for computer assisted transcription and translation

Traditionally, Natural Language Processing (NLP) technologies have mainly focused on full automation. However, full automation often proves unnatural in many applications, where technology is expected to assist rather than replace the human agents. In consequence, Multimodal Interactive (MI) technologies have emerged. On the one hand, the user interactively co-operates with the system to improve system accuracy. On the other hand, multimodality improves system ergonomics. In this paper, we present an application that implements such MI technologies. First, we have designed an Application Programming Interface (API), featuring a client-server framework, to deal with most common NLP MI tasks. Second, we have developed a generic client application. The resulting client-server architecture has been successfully tested with two di erent NLP problems: transcription of text images and translation of texts.

Combination of N-Grams and Stochastic Context-Free Grammars in an Offline Handwritten Recognition System

One area of pattern recognition that is receiving a lot of attention recently is handwritten text recognition. Traditionally, handwritten text recognition systems have been modelled by means of HMM models and n-gram language models. The problem that n-grams present is that they are not able to capture long-term constraints of the sentences. Stochastic context-free grammars (SCFG) can be used to overcome this limitation by rescoring a n-best list generated with the HMM-based recognizer. Howerver, SCFG are known to have problems in the estimation of comlpex real tasks. In this work we propose the use of a combination of n-grams and category-based SCFG together with a word distribution into categories. The category-based approach is thought to simplify the SCFG inference process, while at the same time preserving the description power of the model. The results on the IAM-Database show that this combined scheme outperforms the classical scheme.

A web-based demo to interactive multimodal transcription of historic text images

Paleography experts spend many hours transcribing historic documents, and state-of-the-art handwritten text recognition systems are not suitable for performing this task automatically. In this paper we present the modifications on a previously developed interactive framework for transcription of handwritten text. This system, rather than full automation, aimed at assisting the user with the recognition-transcription process.

CAT-API Framework Prototypes

Traditionally, Pattern Recognition applications have focused on fully automatic systems. However, since their performance is far from being perfect, such automatic systems cannot replace the human expertise. Typically, experts use a Pattern Recognition system in a two-step operation: first, the application generates an output in a fully automatic way; and second, the user revises that output in order to achieve high-quality results. This post-edition approach is rather inefficient and uncomfortable for the user. An alternative, yet effective approach to traditional Pattern Recognition systems is the interactive-predictive paradigm in which both the system is guided by the user and the user is assisted by the system to complete their tasks as efficiently as possible. We present three protypes of Computer Assisted Tools: transcription, translation, and syntactic parsing, respectively. Such prototypes combine the efficiency of the traditional Pattern Recognition systems with the accuracy of the human expertise, enforcing all a multimodal, interactive strategy, and fully integrating the users knowledge into the Pattern Recognition process. User feedback directly allows to improve system accuracy, while multimodality increases both system ergonomy and user acceptability.

A Study on Bilingual Speech Recognition Involving a Minority Language

Multilingual Automatic Speech Recognition (ASR) systems are of great interest in multilingual environments. We have studied the case of the Comunitat Valenciana because the two official languages are Spanish and Valencian. These two languages share most of their phonemes and their syntax and vocabulary are also quite similar since they have influenced each other for many years. In this work, we present the design of the language and the acoustic models for this bilingual situation. Acoustic models can be separate for each language or shared by both of them, and they can be obtained directly from a training corpus or by adapting a previous set of acoustic models. Language models can be separate for each language (monolingual recognition) or mixed for both languages (bilingual recognition). We performed experiments with a small corpus to determine which option was better for this case.

Inference of Stochastic Finite-State Transducers Using N-Gram Mixtures

Statistical pattern recognition has proved to be an interesting framework for machine translation, and stochastic finite-state transducers are adequate models in many language processing areas such as speech translation, computer-assisted translations, etc. The well-known n-gram language models are widely used in this framework for machine translation. One of the application of these n-gram models is to infer stochastic finite-state transducers. However, only simple dependencies can be modelled, but many translations require to take into account strong context and style dependencies. Mixtures of parametric models allow to increase the description power of the statistical models by modelling subclasses of objects. In this work, we propose the use of n-gram mixtures in GIATI, a procedure to infer stochastic finite-state transducers. N-gram mixtures are expected to model topics or writing styles. We present experimental results showing that translation performance can be improved if enough training data is available.