Search

This article provides an overview of the dissemination work carried out in META-NET from 2010 until early 2014; we describe its impact on the regional, national and international level, mainly with regard to politics and the situation of funding for LT topics. This paper documents the initiative’s work throughout Europe in order to boost progress and innovation in our field.

The central objective of the Metanet4u project is to contribute to the establishment of a pan-European digital platform that makes available language resources and services, encompassing both datasets and software tools, for speech and language processing, and supports a new generation of exchange facilities for them.

METANET 4 U is a European project aiming at supporting language technology for European languages and multilingualism. It is a project in the META-NET Network of Excellence, a cluster of projects aiming at fostering the mission of META, which is the Multilingual Europe Technology Alliance, dedicated to building the technological foundations of a multilingual European information society. This paper describe the resources produced at our lab to provide Synthethic voices. Using existing 10h corpus for a male and a female Spanish speakers, voices have been developed to be used in Festival, both with unit-selection and with statistical-based technologies. Furthermore, using data produced for supporting research on intra and inter-lingual voice conversion, four bilingual voices (English/Spanish) have been developed. The paper describes these resources which are available through META. Furthermore, an evaluation is presented to compare different synthesis techniques, influence of amount of data in statistical speech synthesis and the effect of sharing data in bilingual voices

The central objective of the Metanet4u project is to contribute to the establishment of a pan-European digital platform that makes available language resources and services, encompassing both datasets and software tools, for speech and language processing, and supports a new generation of exchange facilities for them.

Deliverable D2.4 del projecte METANET4U (Project CIP #270893)

The central objective of the Metanet4u project is to contribute to the establishment of a pan-European digital platform that makes available language resources and services, encompassing both datasets and software tools, for speech and language processing, and supports a new generation of exchange facilities for them. Peer Reviewed _

This paper describes the BUCEADOR search engine, a web server that allows retrieving. multimedia documents (text, audio, video) in different languages. All the documents are translated into the user language and are presented either as text (for instance, subtitles in video documents) or dubbed audio. The user query consist in a sequence of keywords and can be typed or spoken. Multiple Spoken Language Technologies (SLT) servers have been implemented, such as speech recognition, speech machine translation and text-to-speech conversion. The platform can be used in the four Spanish official (Spanish, Basque, Catalan and Galician) and in English.

The central objective of the Metanet4u project is to contribute to the establishment of a pan-European digital platform that makes available language resources and services, encompassing both datasets and software tools, for speech and language processing, and supports a new generation of exchange facilities for them.

This paper deals with the creation of multiple voices from a Hidden Markov Model based speech synthesis system (HTS). More than 150 Catalan synthetic voices were built using Hidden Markov Models (HMM) and speaker adaptation techniques. Training data for building a Speaker-Independent (SI) model were selected from both a general purpose speech synthesis database (FestCat;) and a database design ed for training Automatic Speech Recognition (ASR) systems (Catalan SpeeCon database). The SpeeCon database was also used to adapt the SI model to different speakers. Using an ASR designed database for TTS purposes provided many different amateur voices, with few minutes of recordings not performed in studio conditions. This paper shows how speaker adaptation techniques provide the right tools to generate multiple voices with very few adaptation data. A subjective evaluation was carried out to assess the intelligibility and naturalness of the generated voices as well as the similarity of the adapted voices to both the original speaker and the average voice from the SI model.

In this paper, multidialectal acoustic modeling based on shar- ing data across dialects is addressed. A comparative study of different methods of combining data based on decision tree clustering algorithms is presented. Approaches evolved differ in the way of evaluating the similarity of sounds between di- alects, and the decision tree structure applied. Proposed systems are tested with Spanish dialects across Spain and Latin Amer- ica. All multidialectal proposed systems improve monodialectal performance using data from another dialect but it is shown that the way to share data is critical. The best combination between similarity measure and tree structure achieves an improvement of 7% over the results obtained with monodialectal systems.

Some Speech Enhancement algorithms based on the iterative Wiener filtering Method due to L1m-Oppenheim [2] are presented. In the original Lim-Oppenheim algorithm, speech AR estimation is carried out using classic second-order analysis, but our algorithms consider a more robust AR modelling. Two different strategies of speech AR estimation are presented and both estimators are trying to see as less amount of noise as possible. First one uses a previous One-Sided Autocorrelation computation, that is a pole-preserving function, and the actual SNR m the second-order LPC analysis is increased. Second one combines advantages of Higher-Order Statistics [1] with a linear combination of AR coefficients, belonging to two consecutive overlapped frames, to assess a less disturbed speech estimation.

This paper introduces a first approach to emotion recognition using RAMSES, the UPC’s speech recognition system. The approach is based on standard speech recognition technology using hidden semi-continuous Markov models. Both the selection of low level features and the design of the recognition system are addressed. Results are given on speaker dependent emotion recognition using the Spanish corpus of INTERFACE Emotional Speech Synthesis Database. The accuracy recognising seven different emotions—the six ones defined in MPEG-4 plus neutral style—exceeds 80% using the best combination of low level features and HMM structure. This result is very similar to that obtained with the same database in subjective evaluation by human judges.

This paper deals with a description of NaniTrans, a tool for segmentation and labeling of speech. The tool is programmed to work on the MATLAB application interface, in any of the supported platforms (Unix, Windows, Macintosh). The tool has been designed to annotate large speech databases, which can be also partially preprocessed (but require manual supervision). It supports the definition of an environment of annotation: set of annotation levels (orthographic, phonetic, etc.), display mode (how to show information), graphic representation (waveform, spectrogram), keyboard short-cuts, etc. This configuration is then used on a speech database. A safe file locking system allows many annotators to work concurrently on the same speech database. The tool is very friendly and easy to use by non experienced annotators, and it is designed to optimize speed using both keyboard and mouse. New options or speech processing tools can be easily added by using any MATLAB or user defined function.

Under the SpeechDat specifications, the Spanish member of SpeechDat consortium has recorded a Catalan database that includes one thousand speakers. This communication describes some experimental work that has been carried out using both the Spanish and the Catalan speech material. A speech recognition system has been trained for the Spanish language using a selection of the phonetically balanced utterances from the 4500 SpeechDat training sessions. Utterances with mispronounced or incomplete words and with intermittent noise were discarded. A set of 26 allophones was selected to account for the Spanish sounds and clustered demiphones have been used as context dependent sub-lexical units. Following the same methodology, a recognition system was trained from the Catalan SpeechDat database. Catalan sounds were described with 32 allophones. Additionally, a bilingual recognition system was built for both the Spanish and Catalan languages. By means of clustering techniques, the suitable set of allophones to cover simultaneously both languages was determined. Thus, 33 allophones were selected. The training material was built by the whole Catalan training material and the Spanish material coming from the Eastern region of Spain (the region where Catalan is spoken). The performance of the Spanish, Catalan and bilingual systems were assessed under the same framework. The Spanish system exhibits a significantly better performance than the rest of systems due to its better training. The bilingual system provides an equivalent performance to that afforded by both language specific systems trained with the Eastern Spanish material or the Catalan SpeechDat corpus.

This paper describes NaniBD, a set of tools designed for transcribing and validating speech databases, developed at the Signal Processing Group (GPS) of the Department of Signal Theory and Communications of the Polytechnic University of Catalonia (TSC/UPC). The main purpose of its development was the need of a revision system in order to validate and annotate the Spanish corpus of SpeechDat (II) in the speech processing environment available at GPS. Despite of this, NaniBD is designed as a general-purpose system that might fit any other database, idiom or speech processing system. So far, the system has been used to revise some 200,000 speech files from three different corpora. In this paper we will focus our attention to the actual implementation used in the transcription of a SpeechDat (II) specifications compatible Catalonian corpus. 1000 speakers, each of them uttering 44 files, compose this corpus. In this application, we use speech-noise detection, automatic recognition of spontaneous prompts, digit and letter to text translation and access to an external database in order to minimise the amount of time spent by human operators in the revision procedure.

We study some speech enhancement algorithms based on the iterative Wiener filtering method due to Lim-Oppenheim [2], where the AR spectral estimation of the speech is carried out using a second-order analysis. But in our algorithms we consider an AR estimation by means of cumulant analysis. This work extends some preceding papers due to the authors, where information of previous speech frames is taken to initiate speech AR modelling of the current frame. Two parameters are introduced to dessign Wiener filter at first iteration of this iterative algorithm. These parameters are the Interframe Factor (IF) and the Previous Frame Iteration (PFI). A detailed study of them shows they allow a very important noise suppression after processing only first iteration of this algorithm, without any appreciable increase of distortion. Finally, the simplest cumulant-based algorithm is applied to Speech Recognition and some preliminary results are presented.

In this paper we are addressed to the problem of speech presence and silences detection. Speech signals possess properties to which higher order statistics are sensitive. Bispectral-based statistics have been proved a good alternative [3] to energy-based statistics in a speech presence detector scheme, but at the cost of computations. Three alternative methods are proposed for a reliable and more computationally efficient detection. Two, lag-and frequency-domain, methods for the computation of the IT and the OT contribution to total skew while keeping the ability of the statistic in [3]. And a method based in the integrated polyspectrum (IP).

The performance of existing speech recognition systems degrades rapidly in the presence of background noise when training and testing cannot be done under the same ambient conditions. The aim of this paper is to propose the application of a simple multilabeling method, instead of the standard vector quantization -so called labeling-, as the front end for a speech recognizer based on the Vector Quantization (VQ) and Hidden Markov Models (HMM) approaches in order to increase its robustness to noise. Furthermore, not only cepstrum but also other features such as energy and dynamic parameters are evaluated and quantized independently in the multilabeling stage to represent more accurately characteristics of speech. The result of this process is a multiple multilabeling. Experimental results in the presence of additive white noise clearly demonstrate its good performance in isolated word recognition in noisy environments.

This communication reports the use of demisyllables for continuous speech recognition in a specific application: the recognition of Spanish numbers. After a brief outline of the recognition system, a description of demisyllable syntactic constraints and one-speaker reference generation is provided. Finally, the recognition performance is assessed by means of two experiments: the recognition of integer numbers from zero to one thousand and telephone numbers uttered in a Spanish way (strings of integers from zero to ninety nine), in both applications the results that the system yielded were excellent.

During the last decade, many efforts have been devoted to the relative importance of associated functions like magnitude and phase of Fourier Transforms in image and signal bandwidth reduction. The reported work deals with the importance of the real envelope and instantaneous frequency in signal analysis/sintesis problems. In this paper authors show a method to parametrize the envelope and instantaneous frequency of a real signal. This method is very closed to spectral analysis methods in the sense that with an appropiate study, time domain and frequency domain can be analyced in a similar way.

This work describes a coder that works at bit rates of 9.6Kbits/s to 32 Kb/ s. Basically the system consist on a waveform DPCM coder with an improvement in the quantizer. This new feature consist on quantizing the prediction error taking into account its waveform characteristics. The prediction residual of a voiced signal is characterized by an energy sychronous with pitch. The envelope of this signal holds this information and can be used to quantize properly the prediction error. In this work a parametric version of the residual envelope is used in two ways: Dynamic bit assignment in time domain and adaptive control of the dynamic range of the quantizer.

In this master thesis we develop a model that surpasses previous studies to be able to detect cyberbullying and other disorders that are a common behaviour in teenagers. We analyze short sentences in social media with new techniques that haven?t been studied in depth in language processing in order to be able to detect these problems. Deep learning is nowadays the common approach for text analysis. However, struggling with dataset size is one of the most common problems. It is not optimal to dedicate thousands of hours to label data by humans every time we want to create a new model. Different techniques have been used over the years to solve or at least minimize this problem, for instance transfer learning or self-learning. One of the most known ways to solve this is by data augmentation. In this thesis we make use of active learning and self-training to address having restrictions of labelled data. We have used data that has not been labeled to improve the performance of our models. The architecture of the model is composed of a Bert model plus a linear layer that projects the Bert sentence embedding into the number of classes we want to detect. We take advantage of this already functional model to label new data that we will use afterwards to create our final model. Using noise techniques we modify the data so the final model has to predict less structured data and learn from difficult scenarios. Thanks to this technique we were able to improve the results in some of the classes, for instance the F-score modified increases by 7% for substance abuse (drugs, alcohol, etc) and 3% in disorders (anxiety, depression and distress) while keeping the performance of the other classes.

This project explores several pipelines to accurately perform text classification in a very specific task, which consists on classifying Catalan and Spanish short texts with very informal language by kids and teenagers. The classification categories are the following abusive topics: aggression and violence, substances, sexuality and disorders (anxiety, depression or distress). We count on a small, highly unbalanced, supervised dataset of approximately 200.000 examples. We first added a robust preprocessing to our database and then we built and tested several pipelines, exploring vectorization (TF-IDF, Doc2Vec) and AI classification techniques (RF, SVM, BERT). The best one was the multilingual version of BERT with our proposed preprocessing without stemming. Another pipeline achieved similar results (surpassing BERT in one of the categories) with a much faster computing time, and that is using a BERT model for extracting embeddings of our short texts and classifying them using SVM. Este proyecto explora varios procesos para realizar con precisión clasificación de texto en una tarea muy específica, que consiste en clasificar textos breves en catalán y español con un lenguaje muy informal escritos por niños y adolescentes. Las categorías de clasificación son los siguientes temas abusivos: agresión y violencia, sustancias, sexualidad y trastornos (ansiedad, depresión o angustia). Contamos con un pequeño conjunto de datos supervisados altamente desequilibrados de aproximadamente 200,000 ejemplos. Primero agregamos un preprocesamiento robusto a nuestra base de datos y luego construimos y probamos varios procesos, explorando la vectorización (TF-IDF, Doc2Vec) y las técnicas de clasificación IA (RF, SVM, BERT). La mejor fue la versión multilingüe de BERT con nuestro preprocesamiento propuesto sin "stemming". Otro proceso logró resultados similares (superando BERT en una de las categorías) con un tiempo de cómputo mucho más rápido, y es la que usó un modelo BERT para extraer "embeddings" de nuestros textos cortos y clasificarlos usando SVM. Aquest projecte explora diversos processos per a realitzar una classificació de textos amb precisió en una tasca molt específica, que consisteix en classificar textos breus en català i castellà amb un llenguatge molt informal escrits per nens i adolescents. Les categories de classificació són les següents: agressió i violència, substàncies, sexualitat i trastorns (ansietat, depressió o angoixa). Comptem amb un petit conjunt de dades supervisat, molt desequilibrat, d'aproximadament 200.000 exemples. Primer vam afegir un preprocessament robust a la nostra base de dades i després vam crear i provar diversos processos, explorant tècniques de vectorització (TF-IDF, Doc2Vec) i de classificació IA (RF, SVM, BERT). El millor procés és la multilingüe de BERT amb la nostra proposta de preprocessament sense "stemming". Un altre procés va obtenir resultats similars (superant BERT en una de les categories) amb un temps de computació molt més ràpid, i és la que utilitza un model BERT per extreure "embeddings" dels nostres textos breus i classificar-los mitjançant SVM.

The following document will drive the reader through the whole development of a project consisting in the investigation of the feasibility and development of an access control security system that uses voice biometrics as its cornerstone to make rooms as safe and controlled as possible in a two-factor authentication approach using open software. For that, an initial decision is presented, in which the technology that is going to be used to make the Speaker Recognition functionality possible is chosen. A detailed explanation of it is included so that the reader will know and understand the basics to follow the project development. With this, the document will go step by step through the different stages that I have faced and the requirements that came up and I had to adapt to, including the integration of the system that is nowadays up and running at IIT. The User Interface of the final application is presented together with an explanation of every step that users will have to follow to use the system correctly. Last, an estimation of the costs associated to this project is presented, which will give the reader an idea of how much starting this project from scratch would cost if the same hardware is used. Also, for those developers that would like to get some of this functionality in their systems, a link to a GitHub repository where they can find the code is also given.