Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 5
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Capítulo o parte de un libro (BookPart)

Active learning for interactive machine translation

  • González Rubio, Jesús
  • Ortiz Martínez, Daniel
  • Casacuberta Nolla, Francisco|||0000-0002-8497-5598
Translation needs have greatly increased during the last years. In many situations, text to be translated constitutes an unbounded stream of data that grows continually with time. An effective approach to translate text documents is to follow an interactive-predictive paradigm in which both the system is guided by the user and the user is assisted by the system to generate error-free translations. Unfortunately, when processing such unbounded data streams even this approach requires an overwhelming amount of manpower. Is in this scenario where the use of active learning techniques is compelling. In this work, we propose different active learning techniques for interactive machine translation. Results show that for a given translation quality the use of active learning allows us to greatly reduce the human effort required to translate the sentences in the stream.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Capítulo o parte de un libro (BookPart)

Multimodal Interactive Parsing

  • Benedí Ruiz, José Miguel
  • Leiva, Luis A.
  • Sánchez Sáez, Ricardo
  • Maca, Mauricio
  • Sánchez Peiró, Joan Andreu|||0000-0003-0423-2020
Probabilistic parsing is a fundamental problem in Computational Linguistics, whose goal is obtaining a syntactic structure associated to a sentence according to a probabilistic grammatical model. Recently, an interactive framework for probabilistic parsing has been introduced, in which the user and the system cooperate to generate error-free parse trees. In an early prototype developed according to this interactive parsing technology, user feedback was provided by means of mouse actions and keyboard strokes. Here we augment the interaction style with support for (non-deterministic) natural handwritten recognition, and provide confidence measures as a visual aid to ease the correction process. Handwriting input seems to be a modality specially suitable for parsing, since the vocabulary size involved in the recognition of syntactic labels is fairly limited and thus intuitively errors should be small. However, errors may increase as handwriting quality (i.e., calligraphy) degrades. To solve this problem, we introduce a late fusion approach that leverages both on-line and off-line information, corresponding to pen strokes and contextual information from the parse trees. We demonstrate that late fusion can effectively help to disambiguate user intention and improve system accuracy., This research has received funding from the EC’s 7th Framework Programme (FP7/2007-13) under grant agreement No.287576- CasMaCat; from the Spanish MEC under the STraDA project (TIN2012-37475- C02-01) and the MITTRAL project (TIN2009-14633-C03-01); from the GV under the Prometeo project; and from the Universidad del Cauca (Colombia)
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Capítulo o parte de un libro (BookPart)

Relevant clouds: leveraging relevance feedback to build tag clouds for image search

  • Leiva Torres, Luis Alberto
  • Villegas Santamaría, Mauricio
  • Paredes Palacios, Roberto
Previous work in the literature has been aimed at exploring tag clouds to improve image search and potentially increase retrieval performance. However, to date none has considered the idea of building tag clouds derived from relevance feedback. We propose a simple approach to such an idea, where the tag cloud gives more importance to the words from the relevant images than the non-relevant ones. A preliminary study with 164 queries inspected by 14 participants over a 30M dataset of automatically annotated images showed that 1) tag clouds derived this way are found to be informative: users considered roughly 20% of the presented tags to be relevant for any query at any time; and 2) the importance given to the tags correlates with user judgments: tags ranked in the first positions tended to be perceived more often as relevant to the topic that users had in mind., Work supported by EU FP7/2007-2013 under grant agreements 600707 (tranScriptorium) and 287576 (CasMaCat), and by the STraDA project (TIN2012-37475-C02-01).
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Capítulo o parte de un libro (BookPart)

Partial least squares for word confidence estimation in machine translation

  • González Rubio, Jesús
  • Navarro Cerdan, José Ramón|||0000-0002-6692-5941
  • Casacuberta Nolla, Francisco|||0000-0002-8497-5598
We present a new technique to estimate the reliability of the words in automatically generated translations. Our approach addresses confidence estimation as a classification problem where a confidence score is to be predicted from a feature vector that represents each translated word. We describe a new set of prediction features designed to capture context information, and propose a model based on partial least squares to perform the classification. Good empirical results are reported in a large-domain news translation task., Work supported by the European Union Seventh Framework Program (FP7/2007-2013) under the CasMaCat project (grants agreement no 287576), by Spanish MICINN under TIASA (TIN2009-14205-C04-02) project, and by the Generalitat Valenciana under grant ALMPR (Prometeo/2009/014).
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Capítulo o parte de un libro (BookPart)

Improving the minimum description length inference of phrase-based translation models

  • Gonzalez Rubio, Jesus
  • Casacuberta Nolla, Francisco|||0000-0002-8497-5598
We study the application of minimum description length (MDL) inference to estimate pattern recognition models for machine translation. MDL is a theoretically-sound approach whose empirical results are however below those of the state-of-the-art pipeline of training heuristics. We identify potential limitations of current MDL procedures and provide a practical approach to overcome them. Empirical results support the soundness of the proposed approach., Work supported by the EU 7th Framework Programme (FP7/2007–2013) under the CasMaCat project (grant agreement no 287576), by Spanish MICINN under grant TIN2012-31723, and by the Generalitat Valenciana under grant ALMPR (Prometeo/2009/014).