To produce high quality translations, human translators need to review and correct machine translation hypothesis in what it is known as post-editing. In order to reduce the human effort of this process, interactive machine translation proposed a collaborative framework in which human and machine work together to generate the translations. Among the many protocols proposed throughout the years, the segment-based one established a paradigm in which the post-editor was allowed to validate correct word sequences from a translation hypothesis and introduced a word correction to help the system improve the next hypothesis. In this work we propose an extension to this protocol: instead of having to the type the complete word correction, the system will complete the user’s correction while they are typing. We evaluated our proposal under a simulated environment, achieving a significant reduction of the human effort.

Abstract Historical documents are an important part of our cultural heritage. Among other task related to their processing, it is important to modernize their language in order to make them accessible to a broader audience and to achieve an orthography consistency to reduce the linguistic variation inherent in them. Language modernization and spelling normalization have those goals in mind. However, they still have a long way to go. Thus, in order to help scholars generate error-free modernizations/normalizations when the quality is essential, we propose an interactive framework based on interactive machine translation. In this work, we deployed two different interactive protocols into these tasks. We evaluated our proposal under simulated environments, observing significant reductions of the human effort.

The quality of the translations generated by Machine Translation (MT) systems has highly improved through the years and but we are still far away to obtain fully automatic high-quality translations. To generate them and translators make use of Computer-Assisted Translation (CAT) tools and among which we find the Interactive-Predictive Machine Translation (IPMT) systems. In this paper and we use bandit feedback as the main and only information needed to generate new predictions that correct the previous translations. The application of bandit feedback reduces significantly the number of words that the translator need to type in an IPMT session. In conclusion and the use of this technique saves useful time and effort to translators and its performance improves with the future advances in MT and so we recommend its application in the actuals IPMT systems.

In this paper, we present a machine translation system implemented by the Translation Centre for the Bodies of the European Union (CdT). The main goal of this project is to create domain-specific machine translation engines in order to support machine translation services and applications to the Translation Centre’s clients. In this article, we explain the entire implementation process of NICE: Neural Integrated Custom Engines. We describe the problems identified and the solutions provided, and present the final results for different language pairs. Finally, we describe the work that will be done on this project in the future.

In the translation industry, human experts usually supervise and post-edit machine translation hypotheses. Adaptive neural machine translation systems, able to incrementally update the underlying models under an online learning regime, have been proven to be useful to improve the efficiency of this workflow. However, this incremental adaptation is somewhat unstable, and it may lead to undesirable side effects. One of them is the sporadic appearance of made-up words, as a byproduct of an erroneous application of subword segmentation techniques. In this work, we extend previous studies on on-the-fly adaptation of neural machine translation systems. We perform a user study involving professional, experienced post-editors, delving deeper on the aforementioned problems. Results show that adaptive systems were able to learn how to generate the correct translation for task-specific terms, resulting in an improvement of the user’s productivity. We also observed a close similitude, in terms of morphology, between made-up words and the words that were expected.

The main aim of the Carabela project was to develop and apply techniques that allow textual searching on massive Spanish collections of 15th-19th century manuscripts. The project focused on a relatively small subset of 125 000 images of collections of interest to underwater archaeology. For this type of manuscripts, state-of-the-art automatic transcription techniques, generally fail to achieve usable transcription accuracy. Therefore, rather than insisting in actual transcription, methodologies for probabilistic indexing of handwritten text images have been adopted. This has allowed us to effectively cope with the intrinsically high degree of uncertainty of the text contained in most historical manuscripts, leading to highly effective systems for textual search and retrieval. Carabela has gone one step further by developing new techniques to classify probabilistically indexed, but otherwise untranscribed, text images according to their textual content. These techniques have been successfully used to automatically classify Carabela bundels (each containing hundreds or thousands of pages) according to their “level of risk” of public exposure, in order to control their access and avoid as much as possible the plundering of Spanish underwater heritage.

We study the application of active learning techniques to the translation of unbounded data streams via interactive neural machine translation.The main idea is to select, from an unbounded stream of source sentences, those worth to be supervised by a human agent.The user will interactively translate those samples.Once validated, these data is useful for adapting the neural machine translation model.We propose two novel methods for selecting the samples to be validated.We exploit the information from the attention mechanism of a neural machine translation system.Our experiments show that the inclusion of active learning techniques into this pipeline allows to reduce the effort required during the process, while increasing the quality of the translation system.Moreover, it enables to balance the human effort required for achieving a certain translation quality.Moreover, our neural system outperforms classical approaches by a large margin.

We propose and study three different novel approaches for tackling the problem of development set selection in Statistical Machine Translation. We focus on a scenario where a machine translation system is leveraged for translating a specific test set, without further data from the domain at hand. Such test set stems from a real application of machine translation, where the texts of a specific e-commerce were to be translated. For developing our development-set selection techniques, we first conducted experiments in a controlled scenario, where labelled data from different domains was available, and evaluated the techniques both with classification and translation quality metrics. Then, the bestperforming techniques were evaluated on the e-commerce data at hand, yielding consistent improvements across two language directions.

The lack of a spelling convention in historical documents makes their orthography to change depending on the author and the time period in which each document was written. This represents a problem for the preservation of the cultural heritage, which strives to create a digital text version of a historical document. With the aim of solving this problem, we propose three approaches—based on statistical, neural and character-based machine translation—to adapt the document’s spelling to modern standards. We tested these approaches in different scenarios, obtaining very encouraging results.

Neural Machine Translation (NMT) has achieved promising results comparable with Phrase-Based Statistical Machine Translation (PBSMT). However, to train a neural translation engine, much more powerful machines are required than those required to develop translation engines based on PBSMT. One solution to reduce the training cost of NMT systems is the reduction of the training corpus through data selection (DS) techniques. There are many DS techniques applied in PBSMT which bring good results. In this work, we show that the data selection technique based on infrequent n-gram occurrence described in (Gasco ́ et al., 2012) commonly used for PBSMT systems also works well for NMT systems. We focus our work on selecting data according to specific corpora using the previously mentioned technique. The specific-domain corpora used for our experiments are IT domain and medical domain. The DS technique significantly reduces the execution time required to train the model between 87% and 93%. Also, it improves translation quality by up to 2.8 BLEU points. The improvements are obtained with just a small fraction of the data that accounts for between 6% and 20% of the total data.

Current automatic machine translation systems require heavy human proofreading to produce high-quality translations.We present a new interactive machine translation approach aimed at providing a natural collaboration between humans and translation systems.As such, we grant the user complete freedom to validate and correct any part of the translations suggested by the system.Our approach is then designed according to the requirements placed by this unrestricted proofreading protocol.In particular, the ability of the system to suggest new translations coherent with the set of potentially disjoint translation segments validated by the user.We evaluate our approach in a usersimulated setting where reference translations are considered the output desired by a human expert.Results show important reductions in the number of edits in comparison to decoupled post-editing and conventional prefix-based interactive translation prediction.Additionally, we provide evidence that it can also reduce the cognitive overload reported for interactive translation systems in previous user studies.

Although traditionally used in the machine translation field, the encoder-decoder framework has been recently applied for the generation of video and image descriptions. The combination of Convolutional and Recurrent Neural Networks in these models has proven to outperform the previous state of the art, obtaining more accurate video descriptions. In this work we propose pushing further this model by introducing two contributions into the encoding stage. First, producing richer image representations by combining object and location information from Convolutional Neural Networks and second, introducing Bidirectional Recurrent Neural Networks for capturing both forward and backward temporal relationships in the input frames.

CASMACAT is a modular, web-based translation workbench that offers advanced functionalities for computer-aided translation and the scientific study of human translation: automatic interaction with machine translation (MT) engines and translation memories (TM) to obtain raw translations or close TM matchesn for conventional post-editing; interactive translation prediction based on an MT engine’s search graph, detailed recording and replay of edit actions and translator’s gaze (the latter via eye-tracking), and the support of e-pen as an alternative input device. The system is open source sofware and interfaces with multiple MT systems.

We present an unsupervised inference procedure for phrase-based translation models based on the minimum description length principle.In comparison to current inference techniques that rely on long pipelines of training heuristics, this procedure represents a theoretically wellfounded approach to directly infer phrase lexicons.Empirical results show that the proposed inference procedure has the potential to overcome many of the problems inherent to the current inference approaches for phrase-based models.

We present the new THOT toolkit for fully-automatic and interactive statistical ma-chine translation (SMT). Initial public ver-sions of THOT date back to 2005 and did only include estimation of phrase-based models. By contrast, the new version of-fers several new features that had not been previously incorporated. The key innova-tions provided by the toolkit are computer-aided translation, including post-editing and interactive SMT, incremental learn-ing and robust generation of alignments at phrase level. In addition to this, the toolkit also provides standard SMT fea-tures such as fully-automatic translation, scalable and parallel algorithms for model training, client-server implementation of the translation functionality, etc. The toolkit can be compiled in Unix-like and Windows platforms and it is released un-der the GNU Lesser General Public Li-cense (LGPL). 1

Current automatic machine translation systems are not able to generate error-free translations and human intervention is often required to correct their output.Alternatively, an interactive framework that integrates the human knowledge into the translation process has been presented in previous works.Here, we describe a new interactive machine translation approach that is able to work with phrase-based and hierarchical translation models, and integrates error-correction all in a unified statistical framework.In our experiments, our approach outperforms previous interactive translation systems, and achieves estimated effort reductions of as much as 48% relative over a traditional post-edition system.

Interactive machine translation (IMT) is an increasingly popular paradigm for semi-automated machine translation, where a human expert is integrated into the core of an automatic machine translation system. The human expert interacts with the IMT system by partially correcting the errors of the system's output. Then, the system proposes a new solution. This process is repeated until the output meets the desired quality. In this scenario, the interaction is typically performed using the keyboard and the mouse. However, speech is also a very interesting input modality since the user does not need to abandon the keyboard to interact with it.

This paper provides the first experimental study of an active learning (AL) scenario for interactive machine translation (IMT). Unlike other IMT implementations where user feedback is used only to improve the predictions of the system, our IMT implementation takes advantage of user feedback to update the statistical models involved in the translation process. We introduce a sentence sampling strategy to select the sentences that are worth to be interactively translated, and a retraining method to update the statistical models with the user-validated translations. Both, the sampling strategy and the retraining process are designed to work in real-time to meet the severe time constraints inherent to the IMT framework. Experiments in a simulated setting showed that the use of AL dramatically reduces user effort required to obtain translations of a given quality.

State-of-the-art approaches to multi-source translation involve a multimodal-like process which applies an individual translation system to each source language. Then, the translations of the individual systems are combined to obtain a consensus output. We propose to use the (generalised) median string as the consensus output of the individual translation systems. Different approximations to the median string are studied as well as different approaches to improve the median string performance when dealing with natural language strings. The proposed approaches were evaluated on the Europarl corpus, achieving significant improvements in translation quality.

Handwritten text is generally captured through two main modalities: off-line and on-line. Smart approaches to handwritten text recognition (HTR) may take advantage of both modalities if they are available. This is for instance the case in computer-assisted transcription of text images, where on-line text can be used to interactively correct errors made by a main off-line HTR system. We present here baseline results on the biMod-IAM-PRHLT corpus, which was recently compiled for experimentation with techniques aimed at solving the proposed multi-modal HTR problem, and is being used in one of the official ICPR-2010 contests.

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 user's 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.

Current machine translation (MT) systems are still not perfect. In practice, the output from these systems needs to be edited to correct errors. A way of increasing the productivity of the whole translation process (MT plus human work) is to incorporate the human correction activities within the translation process itself, thereby shifting the MT paradigm to that of computer-assisted translation. This model entails an iterative process in which the human translator activity is included in the loop: In each iteration, a prefix of the translation is validated (accepted or amended) by the human and the system computes its best (or n-best) translation suffix hypothesis to complete this prefix. A successful framework for MT is the so-called statistical (or pattern recognition) framework. Interestingly, within this framework, the adaptation of MT systems to the interactive scenario affects mainly the search process, allowing a great reuse of successful techniques and models. In this article, alignment templates, phrase-based models, and stochastic finite-state transducers are used to develop computer-assisted translation systems. These systems were assessed in a European project (TransType2) in two real tasks: The translation of printer manuals; manuals and the translation of the Bulletin of the European Union. In each task, the following three pairs of languages were involved (in both translation directions): English-Spanish, English-German, and English-French.

Stochastic finite-state transducers constitute a type of word-based models that allow an easy integration with acoustic model for speech translation. The aim of this work is to develop a novel approach to phrase-based statistical finite-state transducers. In this work, we explore the use of linguistically motivated phrases to build phrase-based models. The proposed phrase-based transducer has been tested and compared to a word-based equivalent machine, yielding promising results in the reported preliminary text and speech translation experiments.

This paper evaluates different approaches on speech to sign language machine translation. The framework of the application focuses on assisting deaf people to apply for the passport or related information. In this context, the main aim is to automatically translate the spontaneous speech, uttered by an officer, into Spanish Sign Language (SSL). In order to get the best translation quality, three alternative techniques have been evaluated: a rule-based approach, a phrase-based statistical approach, and a approach that makes use of stochastic finite state transducers. The best speech translation experiments have reported a 32.0 % SER (Sign Error Rate) and a 7.1 BLEU (BiLingual Evaluation Understudy) including speech recognition errors.

In formal language theory, finite-state transducers are well-know models for simple “input-output” mappings between two languages. Even if more powerful, recursive models can be used to account for more complex mappings, it has been argued that the input-output relations underlying most usual natural language pairs can essentially be modeled by finite-state devices. Moreover, the relative simplicity of these mappings has recently led to the development of techniques for learning finite-state transducers from a training set of input-output sentence pairs of the languages considered. In the last years, these techniques have lead to the development of a number of machine translation systems. Under the statistical statement of machine translation, we overview here how modeling, learning and search problems can be solved by using stochastic finite-state transducers. We also review the results achieved by the systems we have developed under this paradigm. As a main conclusion of this review we argue that, as task complexity and training data scarcity increase, those systems which rely more on statistical techniques tend produce the best results.

Current machine translation systems are far from being perfect. However, such systems can be used in computer-assisted translation to increase the productivity of the (human) translation process. The idea is to use a text-to-text translation system to produce portions of target language text that can be accepted or amended by a human translator using text or speech. These user-validated portions are then used by the text-to-text translation system to produce further, hopefully improved suggestions. There are different alternatives of using speech in a computer-assisted translation system: From pure dictated translation to simple determination of acceptable partial translations by reading parts of the suggestions made by the system. In all the cases, information from the text to be translated can be used to constrain the speech decoding search space. While pure dictation seems to be among the most attractive settings, unfortunately perfect speech decoding does not seem possible with the current speech processing technology and human error-correcting would still be required. Therefore, approaches that allow for higher speech recognition accuracy by using increasingly constrained models in the speech recognition process are explored here. All these approaches are presented under the statistical framework. Empirical results support the potential usefulness of using speech within the computer-assisted translation paradigm.

Statistical techniques and grammatical inference have been used for dealing with automatic speech recognition with success, and can also be used for speech-to-speech machine translation. In this paper, new advances on a method for finite-state transducer inference are presented. This method has been tested experimentally in a speech-input translation task using a recognizer that allows a flexible use of models by means of efficient algorithms for on-thefly transducer composition. These are the first reported results of a speech-to-speech translation task involving European Portuguese input that we know of.

In this paper, we present the Thot toolkit, a set of tools to train phrase-based models for statistical machine translation, which is publicly available as open source software. The toolkit obtains phrase-based models from word-based alignment models; to our knowledge, this functionality has not been offered by any publicly available toolkit. The Thot toolkit also implements a new way for estimating phrase models, this allows to obtain more complete phrase models than the methods described in the literature, including a segmentation length submodel. The toolkit output can be given in different formats in order to be used by other statistical machine translation tools like Pharaoh, which is a beam search decoder for phrase-based alignment models which was used in order to perform translation experiments with the generated models. Additionally, the Thot toolkit can be used to obtain the best alignment between a sentence pair at phrase level.

Probabilistic finite-state machines are used today in a variety of areas in pattern recognition, or in fields to which pattern recognition is linked: computational linguistics, machine learning, time series analysis, circuit testing, computational biology, speech recognition, and machine translation are some of them. In Part I of this paper, we survey these generative objects and study their definitions and properties. In Part II, we will study the relation of probabilistic finite-state automata with other well-known devices that generate strings as hidden Markov models and n-grams and provide theorems, algorithms, and properties that represent a current state of the art of these objects.

Probabilistic finite-state machines are used today in a variety of areas in pattern recognition or in fields to which pattern recognition is linked. In part I of this paper, we surveyed these objects and studied their properties. In this part, we study the relations between probabilistic finite-state automata and other well-known devices that generate strings like hidden Markov models and n-grams and provide theorems, algorithms, and properties that represent a current state of the art of these objects.

We propose a statistical approach to speech-to-speech translation that uses finite-state models in all levels. Acoustic hidden Markov models (HMMs) model the pronunciation of the input-language phonemes and words, while the input–output word mapping, along with the syntax of the output language, are jointly modeled by means a large stochastic finite-state transducer. This allows for a complete integration of all the models so that the translation process can be performed by searching for an optimal path of states through the integrated network. As in speech recognition, HMMs can be trained from an input-language speech corpus, and the translation model is learned automatically from a parallel (text) training corpus. This approach has been assessed in the framework of the EuTrans project, funded by the European Union. Extensive experiments have been carried out with speech-input translations from Spanish to English and from Italian to English in applications involving the interaction (by telephone) of a customer with the front desk of a hotel. A summary of the most relevant results is presented.

Finite-state transducers are models that are being used in different areas of pattern recognition and computational linguistics. One of these areas is machine translation, in which the approaches that are based on building models automatically from training examples are becoming more and more attractive. Finite-state transducers are very adequate for use in constrained tasks in which training samples of pairs of sentences are available. A technique for inferring finite-state transducers is proposed in this article. This technique is based on formal relations between finite-state transducers and rational grammars. Given a training corpus of source-target pairs of sentences, the proposed approach uses statistical alignment methods to produce a set of conventional strings from which a stochastic rational grammar (e.g., an n-gram) is inferred. This grammar is finally converted into a finite-state transducer. The proposed methods are assessed through a series of machine translation experiments within the framework of the E u Trans project.

The interpretation of handwritten sentences is carried out using a holistic approach in which both text image recognition and the interpretation itself are tightly integrated. Conventional approaches follow a serial, first-recognition then-interpretation scheme which cannot adequately use semantic–pragmatic knowledge to recover from recognition errors. Stochastic finite-sate transducers are shown to be suitable models for this integration, permitting a full exploitation of the final interpretation constraints. Continuous-density hidden Markov models are embedded in the edges of the transducer to account for lexical and morphological constraints. Robustness with respect to stroke vertical variability is achieved by integrating tangent vectors into the emission densities of these models. Experimental results are reported on a syntax-constrained interpretation task which show the effectiveness of the proposed approaches. These results are also shown to be comparatively better than those achieved with other conventional, N-gram-based techniques which do not take advantage of full integration.

The goal of the AMETRA project is to make a computer-assisted translation tool from the Spanish language to the Basque language under the memory-based translation framework. The system is based on a large collection of bilingual word-segments. These segments are obtained using linguistic or statistical techniques from a Spanish-Basque bilingual corpus consisting of sentences extracted from the Basque Country’s of£cial government record. One of the tasks within the global information document of the AMETRA project is to study the combination of well-known statistical techniques for the translation of short sequences and techniques for memory-based translation. In this paper, we address the problem of constructing a statistical module to deal with the task of translating segments. The task undertaken in the AMETRA project is compared with other existing translation tasks, This study includes the results of some preliminary experiments we have carried out using well-known statistical machine translation tools and techniques.

We address the problem of smoothing the probability distribution defined by a finite state automaton. Our approach extends the ideas employed for smoothing n-gram models. This extension is obtained by interpreting n-gram models as finite state models. The experiments show that our smoothing improves perplexity over smoothed n-grams and Error Correcting Parsing techniques.

The problem of cyclic sequence alignment is considered. Most existing optimal methods for comparing cyclic sequences are very time consuming. For applications where these alignments are intensively used, optimal methods are seldom a feasible choice. The alternative to an exact and costly solution is to use a close-to-optimal but cheaper approach. In previous works, we have presented three suboptimal techniques inspired on the quadratic-time suboptimal algorithm proposed by Bunke and Bühler. Do these approximate approaches come sufficiently close to the optimal solution, with a considerable reduction in computing time? Is it thus worthwhile investigating these approximate methods? This paper shows that approximate techniques are good alternatives to optimal methods.

Typically, the lexicon models used in statistical machine translation systems do not include any kind of linguistic or contextual information, which often leads to problems in performing a correct word sense disambiguation. One way to deal with this problem within the statistical framework is to use maximum entropy methods. In this paper, we present how to use this type of information within a statistical machine translation system. We show that it is possible to significantly decrease training and test corpus perplexity of the translation models. In addition, we perform a rescoring of N-Best lists using our maximum entropy model and thereby yield an improvement in translation quality. Experimental results are presented on the so-called "Verbmobil Task".

Nowadays, the most successful speech recognition systems are based on stochastic finite-state networks (hidden Markov models and n-grams). Speech translation can be accomplished in a similar way as speech recognition. Stochastic finite-state transducers, which are specific stochastic finite-state networks, have proved very adequate for translation modeling. In this work a speech-to-speech translation system, the EuTRANS system, is presented. The acoustic, language and translation models are finite-state networks that are automatically learnt from training samples. This system was assessed in a series of translation experiments from Spanish to English and from Italian to English in an application involving the interaction (by telephone) of a customer with a receptionist at the front-desk of a hotel.

The increasing interest in the statistical approach to Machine Translation is due to the development of effective algorithms for training the probabilistic models proposed so far. However, one of the open problems with statistical machine translation is the design of efficient algorithms for translating a given input string. For some interesting models, only (good) approximate solutions can be found. Recently, a dynamic programming-like algorithm for the IBM-Model 2 has been proposed which is based on an iterative process of refinement solutions. A new dynamic programming-like algorithm is proposed here to deal with more complex IBM models (models 3 to 5). The computational cost of the algorithm is reduced by using an alignment-based pruning technique. Experimental results with the so-called “Tourist Task” are also presented.

A finite-state, rule-based morphological analyser is presented here, within the framework of machine translation system TAVAL. This morphological analyser introduces specific features which are particularly useful for translation, such as the detection and morphological tagging of word groups that act as a single lexical unit for translation purposes. The case where words in one such group are not strictly contiguous is also covered. A brief description of the Spanish-to-Catalan and Catalan-to-Spanish translation system TAVAL is given in the paper.

A new system for statistical natural language translation for languages with similar grammar is introduced. Specifically, it can be used with Romanic Languages, such as French, Spanish or Catalan. The statistical translation uses two sources of information: a language model and a translation model. The language model used is a standard trigram model. A new approach is defined in the translation model. The two main properties of the translation model are: the translation probabilities are computed between groups of words and the alignment between those groups is monotone. That is, the order between the word groups in the source sentence is conserved in the target sentence. Once, the translation model has been defined, we present an algorithm to infer its parameters from training samples. The translation process is carried out with an efficient algorithm based on stack-decoding. Finally, we present some translation results from Catalan to Spanish and compare our model with other conventional models.

Statistical pattern recognition is a promising framework for text-to-text translation. However, a natural extension to speech-input translation is not straightforward. We present a method to deal with the speech input statistical translation problem that could be considered as a step towards a fully integrated recognition-translation procedure. In this version a word graph was used in the input as a representation of the acoustic of a given utterance. As a case study, experimental results with the so-called "Traveller task" are presented by using a text-input statistical translator.

A string that minimizes the sum of distances to the strings of a given set is known as (generalized) median string of the set. This concept is important in pattern recognition for modelling a (large) set of garbled strings or patterns. The search of such a string is an NP-Hard problem and, therefore, no efficient algorithms to compute the median strings can be designed. A greedy approach has been proposed to compute an approximate median string of a set of strings. In this work an algorithm is proposed that iteratively improves the approximate solution given above. Experiments have been carried out on synthetic and real data to compare the performances of the approximate median string with the conventional set median. These experiments showed that the proposed median string is a better representation of a given set than the corresponding set median.

Current speech technology allows us to build efficient speech recognition systems. However, model learning of knowledge sources in a speech recognition system is not a closed problem. In addition, lower demand of computational requirements are crucial to building real-time systems. ATROS is an automatic speech recognition system whose acoustic, lexical, and syntactical models can be learnt automatically from training data by using similar techniques. In this paper, an improved version of ATROS which can deal with large smoothed language models and with large vocabularies is presented. This version supports acoustic and syntactical models trained with advanced grammatical inference techniques. It also incorporates new data structures and improved search algorithms to reduce the computational requirements for decoding. The system has been tested on a Spanish task of queries to a geographical database (with a vocabulary of 1,208 words).

In this paper, we introduce several hybrid connectionist-structural acoustic models for contextindependent phone-like units in the atros recognition system. The structural part of the acoustic models has been modeled with Markov chains, and a multilayer perceptron (or a committee of multilayer perceptrons) is used to estimate the emission probabilities of the Markov chains. We compare the recognition performance attained by these models with the performance obtained by classical continuous density hidden Markov models on a semantic restricted task. 1 Introduction Acoustic phonetic-decoding for continuous speech recognition is an open problem in speech research, because the nal performance of an automatic speech recognition system greatly depends on the acoustic modeling quality. Hidden Markov models (HMMs) of phone-like units are the most popular option for modeling speech sounds. Under the statistical framework [1], the problem of speech recognition is to search for a word string ^ ...

The increasing interest in the statistical approach to Machine Translation is due to the development of effective algorithms for training the probabilistic models proposed so far. However, one of the open problems with Statistical Machine Translation is the design of efficient algorithms for translating a given input string. For some interesting models, only (good) approximate solutions can be found. Recently a Dynamic Programming-like algorithm has been introduced which computes approximate solutions for some models. These solutions can be improved by using an iterative algorithm that refines the succesive solutions and uses a smoothing technique for some probabilistic distribution of the models based on an interpolation of different distributions. The technique resulting from this combination has been tested on the "Tourist Task" corpus, which was generated in a semi-automated way. The best results achieved were a word-error rate of 9.3% and a sentence-error rate of 44.4%.

This paper extends previous work exploring the use of Subsequential Transducers to perform speech-input translation in limited-domain tasks. This is done following an integrated approach in which a Subsequential Transducer replaces the input-language model of a conventional speech recognition system, and is used both as language and translation model. This way, the search for the recognised sentence also produces the corresponding translation. A corpus-based approach is adopted in order to build the required models from training data. Experimental results are presented for the translation task considered in the EUTRANS project: one in the hotel domain with more than 500 words per language and language perplexities near to 10.

Stochastic Grammars are the most usual models in Syntactic Pattern Recognition. Both components of a Stochastic Grammar, the characteristic grammar and the probabilities attached to the rules, can be learnt automatically from training samples. In this paper, first a review of some algorithms are presented to infer the probabilistic component of Stochastic Regular and Context-Free Grammars under the framework of the Growth Transformations. On the other hand, with Stochastic Grammars, the patterns must be represented as strings over a finite set of symbols. However, the most natural representation in many Syntactic Pattern Recognition applications (i.e. speech) is as sequences of vectors from a feature vector space, that is, a continuous representation. Therefore, to obtain a discrete representation of the patterns, some quantization errors are introduced in the representation process. To avoid this drawback, a formal presentation of a semi-continuous extension of the Stochastic Regular and Context-Free Grammars is studied and probabilistic estimation algorithms are developed in this paper. In this extension, sequences of vectors, instead of strings of symbols, can be processed with Stochastic Grammars.

The design of current acoustic-phonetic decoders for a specific language involves the selection of an adequate set of sublexical units, and a choice of the mathematical framework for modelling the corresponding units. In this work, the baseline chosen for continuous Spanish speech consists of 23 sublexical units that roughly correspond to the 24 Spanish phonemes. The process of selection of such a baseline was based on language phonetic criteria and some experiments with an available speech corpora. On the other hand, two types of models were chosen for this work, conventional Hidden Markov Models and Inferred Stochastic Regular Grammars. With these two choices we could compare classical Hidden Markov modelling where the structure of a unit-model is deductively supplied, with Grammatical Inference modelling where the baseforms of model-units are automatically generated from training samples. The best speaker-independent phone recognition rate was 64% for the first type of modelling, and 66% for the second type.

The authors propose obtaining the structure of phonetic units from training samples of speech automatically by using two specific grammatical inference (GI) algorithms: the error correcting GI algorithm and the morphic generator GI (MGGI) methodology. They describe the adequacy of the properties and capabilities of both methods for the modeling of subword units of speech (such as phonemes). They also report preliminary results obtained in their application to a continuous speech recognition, task. The results obtained with the semicontinuous MGGI methodology are shown to be very encouraging and can be improved with the use of some phonological grammar.>

A technique is described in which all the components of a hidden Markov model are learnt from training speech data. The structure or topology of the model (i.e. the number of states and the actual transitions) is obtained by means of an error-correcting grammatical inference algorithm (ECGI). This structure is then reduced by using an appropriate state pruning criterion. The statistical parameters that are associated with the obtained topology are estimated from the same training data by means of the standard Baum-Welch algorithm. Experimental results showing the applicability of this technique to speech recognition are presented.>

In the literature on automatic speech recognition, the popular hidden Markov models (HMMs), left-to-right hidden Markov models (LRHMMs), Markov source models (MSMs), and stochastic regular grammars (SRGs) are often proposed as equivalent models. However, no formal relations seem to have been established among these models to date. A study of these relations within the framework of formal language theory is presented. The main conclusion is that not all of these models are equivalent, except certain types of hidden Markov models with observation probability distribution in the transitions, and stochastic regular grammar.>

A new framework is introduced which allows the formulation of difficult structural classification tasks in terms of decision-theoretic-based pattern recognition. It is based on extending the classical formulation of generalized linear discriminant functions so as to permit each given object to have a different vector representation in each class. The proposed extension properly accounts for the corresponding extension of the classical learning techniques of linear discriminant functions in a way such that the convergence of the extended techniques can still be proved. The proposed framework can be considered as a hybrid methodology in which both structural and decision-theoretic pattern recognition are integrated. Furthermore, it can be considered as a means to achieve convenient tradeoffs between the inductive and deductive ways of knowledge acquisition, which can result in rendering tractable the possibly hard original inductive learning problem associated with the given task. The proposed framework and methods are illustrated through their use in two difficult structural classification tasks, showing both the appropriateness and the capability of these methods to obtain useful results.

A methodology is proposed for the inference of regular grammars from positive samples of their languages. It is mainly based on the generative mechanism associated with local languages, which allows us to obtain arbitrary regular languages by applying morphic operators to local languages. The actual inference procedure of this methodology consists of obtaining a local language associated with the given positive sample. This procedure, which is very simple, is always the same, regardless of the problem considered, while the task-dependent features that are desired for the inferred languages, are specified through the definition of certain task-appropriate symbol renaming functions (morphisms).