Pasar al contenido principal

PROCESADO DE AUDIO, HABLA Y LENGUAJE PARA ANALISIS DE INFORMACION MULTIMEDIA - UZ

TIN2014-54288-C4-2-R

•
Nombre agencia financiadora Ministerio de Economía y Competitividad
Acrónimo agencia financiadora MINECO
Programa Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Convocatoria Proyectos I+D+i Retos de Investigación
Año convocatoria 2014
Plan Estatal 2013-2016 Plan Estatal de Investigación Científica y Técnica y de Innovación
Unidad de gestión Secretaría de Estado de Educación, Formación Profesional y Universidades
Centro beneficiario UNIVERSIDAD DE ZARAGOZA
Identificador persistente http://dx.doi.org/10.13039/501100003329

Resultados relacionados

Resultados totales (Incluyendo duplicados): 3
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

ASLP-MULAN: Audio speech and language processing for multimedia analytics, ASLP-MULAN: Procesado de audio, habla y lenguaje para análisis de información multimedia

  • Ferreiros Lopez, Javier
  • Pardo Muñoz, José Manuel
  • Segarra Soriano, Encarnación
  • Ortega Giménez, Alfonso
  • Lleida, Eduardo
  • Torres, María Inés
  • Justo, Raquel
  • Hurtado Oliver, Lluis Felip|||0000-0002-1877-0455
[EN] Our intention is generating the right mixture of audio, speech and language technologies with big data ones. Some audio, speech and language automatic technologies are available or gaining enough degree of maturity as to be able to help to this objective: automatic speech transcription, query by spoken example, spoken information retrieval, natural language processing, unstructured multimedia contents transcription and description, multimedia files summarization, spoken emotion detection and sentiment analysis, speech and text understanding, etc. They seem to be worthwhile to be joined and put at work on automatically captured data streams coming from several sources of information like YouTube, Facebook, Twitter, online newspapers, web search engines, etc. to automatically generate reports that include both scientific based scores and subjective but relevant summarized statements on the tendency analysis and the perceived satisfaction of a product, a company or another entity by the general population., [ES] Nuestra intención es generar la mezcla ideal de tecnologías del audio, el habla y el lenguaje con las de big data. Algunas tecnologías automáticas del procesado de audio, habla y lenguaje están adquiriendo suficiente grado de madurez para ser capaces de ayudar a este objetivo: transcripción automática del habla, métodos de búsqueda por habla, recuperación de documentos hablados, procesado del lenguaje natural, transcripción y descripción de contenidos multimedia no estructurados, resumen de ficheros multimedia, detección de emoción en el habla y análisis de sentimientos, comprensión de texto y habla, etc. Parece que merece la pena unirlos y ponerlos a trabajar sobre secuencias de datos obtenidos automáticamente procedentes de diversas fuentes de información como YouTube, Facebook, Twitter, periódicos digitales, buscadores de internet, etc. para generar informes que incluyan tanto puntuaciones basadas en análisis cuantitativo como expresiones resumidas subjetivas pero significativas sobre el análisis de tendencias y la satisfacción percibida sobre un producto, una empresa u otra entidad., This Project is founded by the “Ministerio de Economía y Competitividad” TIN2014-54288-C4 and there are four reseach groups involved: ELiRF (Universitat Politècnica de València), ViVoLab (Universidad de Zaragoza), SPIN (Universidad del Pais Vasco), GTH (Universidad Politécnica de Madrid).
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publicación Artículo científico (JournalArticle)

Error correction in lightly supervised alignment of broadcast subtitles

  • Olcoz, J.
  • Saz, O.
  • Hain, T.
This paper presents a range of error correction techniques aimed at improving the accuracy of a lightly supervised alignment task for broadcast subtitles. Lightly supervised approaches are frequently used in the multimedia domain, either for subtitling purposes or for providing a more reliable source for training speech based systems. The proposed methods focus on directly correcting of the alignment output using different techniques to infer word insertions and words with inaccurate time boundaries. The features used by the classification models are the outputs from the alignment system, such as confidence measures, and word or segment duration. Experiments in this paper are based on broadcast material provided by the BBC to the Multi Genre Broadcast (MGB) challenge participants. Results, show that the order alignment F measure improves up to 2.6% absolute (15.8% relative) when combining insertion and word boundary correction.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publicación Artículo científico (JournalArticle)

Tied hidden factors in neural networks for end-To-end speaker recognition

  • Miguel, A.
  • Llombart, J.
  • Ortega, A.
  • Lleida, E.
In this paper we propose a method to model speaker and session variability and able to generate likelihood ratios using neural networks in an end-To-end phrase dependent speaker verification system. As in Joint Factor Analysis, the model uses tied hidden variables to model speaker and session variability and a MAP adaptation of some of the parameters of the model. In the training procedure our method jointly estimates the network parameters and the values of the speaker and channel hidden variables. This is done in a two-step backpropagation algorithm, first the network weights and factor loading matrices are updated and then the hidden variables, whose gradients are calculated by aggregating the corresponding speaker or session frames, since these hidden variables are tied. The last layer of the network is defined as a linear regression probabilistic model whose inputs are the previous layer outputs. This choice has the advantage that it produces likelihoods and additionally it can be adapted during the enrolment using MAP without the need of a gradient optimization. The decisions are made based on the ratio of the output likelihoods of two neural network models, speaker adapted and universal background model. The method was evaluated on the RSR2015 database.