Luděk Müller

Automatic online subtitling of the czech parliament meetings

This paper describes a LVCSR system for automatic online subtitling (closed captioning) of TV transmissions of the Czech Parliament meetings. The recognition system is based on Hidden Markov Models, lexical trees and bigram language model. The acoustic model is trained on 40 hours of parliament speech and the language model on more than 10M tokens of parliament speech trancriptions. The first part of the article is focused on text normalization and class-based language model preparation. The second part describes the recognition network and its decoding with respect to real-time operation demands using up to 100k vocabulary. The third part outlines the application framework allowing generation and displaying of subtitles for any audio/video source. Finally, experimental results obtained on parliament speeches with recognition accuracy varying from 80 to 95 % (according to the discussed topic) are reported and discussed.

Adaptation of a feedforward artificial neural network using a linear transform

In this paper we present a novel method for adaptation of a multi-layer perceptron neural network (MLP ANN). Nowadays, the adaptation of the ANN is usually done as an incremental retraining either of a subset or the complete set of the ANN parameters. However, since sometimes the amount of the adaptation data is quite small, there is a fundamental drawback of such approach - during retraining, the network parameters can be easily overfitted to the new data. There certainly are techniques that can help overcome this problem (early-stopping, cross-validation), however application of such techniques leads to more complex and possibly more data hungry training procedure. The proposed method approaches the problem from a different perspective. We use the fact that in many cases we have an additional knowledge about the problem. Such additional knowledge can be used to limit the dimensionality of the adaptation problem. We applied the proposed method on speaker adaptation of a phoneme recognizer based on TRAPS (Temporal Patterns) parameters. We exploited the fact that the employed TRAPS parameters are constructed using log-outputs of mel-filter bank and by virtue of reformulating the first layer weight matrix adaptation problem as a mel-filter bank output adaptation problem, we were able to significantly limit the number of free variables. Adaptation using the proposed method resulted in a substantial improvement of phoneme recognizer accuracy.

Negative association between plasma levels of adiponectin and polychlorinated biphenyl 153 in obese women under non-energy-restrictive regime

The aim of this study was to reveal whether accumulation of the persistent organic pollutants (POPs), especially polychlorinated biphenyl (2,2′,4,4′,5,5′-hexachlorobiphenyl, PCB 153), affects plasma levels of adiponectin in obese patients. The study was designed as a longitudinal intervention trial with a control group, where 27 obese women (body mass index (BMI)>30 kg/m2; age 21–74 years) were studied before (OB) and after (OB-LCD) a 3-month low-calorie-diet intervention (LCD; 5 MJ daily). As the control group, 9 female volunteers without LCD intervention were used (C; BMI=19–25 kg/m2; age 21–64 years). Plasma levels of PCB 153 were measured by high-resolution gas chromatography with electron capture detection; total adiponectin and insulin plasma levels were quantified by immunoassays; and adiponectin multimeric complexes were quantified by immunoblotting. Plasma levels of total adiponectin, high and medium molecular weight multimers significantly negatively correlated with plasma levels of PCB 153 in OB, but not in C or in OB-LCD, whereas the LCD intervention lowered BMI by 3.3±3.0 kg/m2. Our results may suggest suppression of adiponectin by PCB 153 in obese women under non-energy-restrictive regime, which may contribute to the known association of PCB 153 and other POPs with type 2 diabetes.

Benefit of proper language processing for Czech speech retrieval in the CL-SR task at CLEF 2006

The paper describes the system built by the team from the University of West Bohemia for participation in the CLEF 2006 CL-SR track. We have decided to concentrate only on the monolingual searching in the Czech test collection and investigate the effect of proper language processing on the retrieval performance. We have employed the Czech morphological analyser and tagger for that purposes. For the actual search system, we have used the classical tf.idf approach with blind relevance feedback as implemented in the Lemur toolkit. The results indicate that a suitable linguistic preprocessing is indeed crucial for the Czech IR performance.

Automatic lemmatizer construction with focus on OOV words lemmatization

This paper deals with the automatic construction of a lemmatizer from a Full Form – Lemma (FFL) training dictionary and with lemmatization of new, in the FFL dictionary unseen, i.e. out-of-vocabulary (OOV) words. Three methods of lemmatization of three kinds of OOV words (missing full forms, unknown words, and compound words) are introduced. These methods were tested on Czech test data. The best result (recall: 99.3 % and precision: 75.1 %) has been achieved by a combination of these methods. The lexicon-free lemmatizer based on the method of lemmatization of unknown words (lemmatization patterns method) is introduced too.

Application of LSTM Neural Networks in Language Modelling

Artificial neural networks have become state-of-the-art in the task of language modelling on a small corpora. While feed-forward networks are able to take into account only a fixed context length to predict the next word, recurrent neural networks (RNN) can take advantage of all previous words. Due the difficulties in training of RNN, the way could be in using Long Short Term Memory (LSTM) neural network architecture.

Initialization of fMLLR with sufficient statistics from similar speakers

One of the most utilized adaptation techniques is the feature Maximum Likelihood Linear Regression (fMLLR). In comparison with other adaptation methods the number of free parameters to be estimated significantly decreases. Thus, the method is well suited for situations with small amount of adaptation data. However, fMLLR still fails in situations with extremely small data sets. Such situations can be solved through proper initialization of fMLLR estimation adding some a-priori information. In this paper a novel approach is proposed solving the problem of fMLLR initialization involving statistics from speakers acoustically close to the speaker to be adapted. Proposed initialization suitably substitutes missing adaptation data with similar data from a training database, fMLLR estimation becomes well-conditioned, and the accuracy of the recognition system increases even in situations with extremely small data sets.

Czech text-to-sign speech synthesizer

Recent research progress in developing of the Czech - Sign Speech synthesizer is presented. The current goal is to improve the system for automatic synthesis to produce accurate synthesis of the Sign Speech. The synthesis system converts written text to an animation of an artificial human model (avatar). This includes translation of text to sign phrases and their conversion to the animation of the avatar. The animation is composed of movements and deformations of segments of hands, a head and also a face. The system has been evaluated by two initial perceptual tests. The perceptual tests indicate that the designed synthesis system is capable to produce the intelligible Sign Speech.

Czech-Sign speech corpus for semantic based machine translation

This paper describes progress in a development of the human-human dialogue corpus for machine translation of spoken language. We have chosen a semantically annotated corpus of phone calls to a train timetable information center. The phone calls consist of inquiries regarding their train traveler plans. Corpus dialogue act tags incorporate abstract semantic meaning. We have enriched a part of the corpus with Sign Speech translation and we have proposed methods how to do automatic machine translation from Czech to Sign Speech using semantic annotation contained in the corpus.

Using the Lemmatization technique for phonetic transcription in text-to-speech system

This paper deals with a lemmatization technique and its using for phonetic transcription of exceptional words. The lemmatizer is based on language morphology and uses a lexicon of basic word forms and a set of inversion derivation rules to acquire lemmatization rules, which are essential for finding word bases. The lemmatization algorithm and its necessary modifications for transcription of exceptional words are described. The main goal of the designed system is to save computer memory for exceptional lexicon storing. The experimental results showed that it is possible to save from 18.3% (English) to 98.4% (Finnish) of the full lexicon size. Hence, the described technique can be applied with advantage for high inflectional and agglutinative languages.