In this paper, we present TED-LIUM release 3 corpus 3 dedicated to speech recognition in English, which multiplies the available data to train acoustic models in comparison with TED-LIUM 2, by a factor of more than two. We present the recent development on Automatic Speech Recognition (ASR) systems in comparison with the two previous releases of the TED-LIUM Corpus from 2012 and 2014. We demonstrate that, passing from 207 to 452 hours of transcribed speech training data is really more useful for end-to-end ASR systems than for HMM-based state-of-the-art ones. This is the case even if the HMMbased ASR system still outperforms the end-to-end ASR system when the size of audio training data is 452 hours, with a Word Error Rate (WER) of 6.7% and 13.7%, respectively. Finally, we propose two repartitions of the TED-LIUM release 3 corpus: the legacy repartition that is the same as that existing in release 2, and a new repartition, calibrated and designed to make experiments on speaker adaptation. Similar to the two first releases, TED-LIUM 3 corpus will be freely available for the research community.

Named entity recognition (NER) is among SLU tasks that usually extract semantic information from textual documents. Until now, NER from speech is made through a pipeline process that consists in processing first an automatic speech recognition (ASR) on the audio and then processing a NER on the ASR outputs. Such approach has some disadvantages (error propagation, metric to tune ASR systems sub-optimal in regards to the final task, reduced space search at the ASR output level,...) and it is known that more integrated approaches outperform sequential ones, when they can be applied. In this paper, we present a first study of end-to-end approach that directly extracts named entities from speech, though a unique neural architecture. On a such way, a joint optimization is able for both ASR and NER. Experiments are carried on French data easily accessible, composed of data distributed in several evaluation campaign. Experimental results show that this end-to-end approach provides better results (F-measure=0.69 on test data) than a classical pipeline approach to detect named entity categories (F-measure=0.65).

This paper addresses the problem of automatic speech recognition (ASR) error detection and their use for improving spoken language understanding (SLU) systems. In this study, the SLU task consists in automatically extracting, from ASR transcriptions, semantic concepts and concept/values pairs in a e.g touristic information system. An approach is proposed for enriching the set of semantic labels with error specific labels and by using a recently proposed neural approach based on word embeddings to compute well calibrated ASR confidence measures. Experimental results are reported showing that it is possible to decrease significantly the Concept/Value Error Rate with a state of the art system, outperforming previously published results performance on the same experimental data. It also shown that combining an SLU approach based on conditional random fields with a neural encoder/decoder attention based architecture, it is possible to effectively identifying confidence islands and uncertain semantic output segments useful for deciding appropriate error handling actions by the dialogue manager strategy.

This paper focuses on error detection in Automatic Speech Recognition (ASR) outputs. A neural network architecture is proposed, which is well suited to handle continuous word representations, like word embeddings. In a previous study, the authors explored the use of linguistic word embeddings, and more particularly their combination. In this new study, the use of acoustic word embeddings is explored. Acoustic word embeddings offer the opportunity of an a priori acoustic representation of words that can be compared, in terms of similarity, to an embedded representation of the audio signal. First, we propose an approach to evaluate the intrinsic performances of acoustic word embeddings in comparison to orthographic representations in order to capture discriminative phonetic information. Since French language is targeted in experiments, a particular focus is made on homophone words. Then, the use of acoustic word embeddings is evaluated for ASR error detection. The proposed approach gets a classification error rate of 7.94% while the previous state-of-the-art CRFbased approach gets a CER of 8.56% on the outputs of the ASR system which won the ETAPE evaluation campaign on speech recognition of French broadcast news.

Modern Natural Language Processing (NLP) makes intensive use of deep learning methods because of the accuracy they offer for a variety of applications. Due to the significant environmental impact of deep learning, costbenefit analysis including carbon footprint as well as accuracy measures has been suggested to better document the use of NLP methods for research or deployment. In this paper, we review the tools that are available to measure energy use and CO 2 emissions of NLP methods. We describe the scope of the measures provided and compare the use of six tools (carbon tracker, experiment impact tracker, green algorithms, ML CO2 impact, energy usage and cumulator) on named entity recognition experiments performed on different computational set-ups (local server vs. computing facility). Based on these findings, we propose actionable recommendations to accurately measure the environmental impact of NLP experiments.

No abstract

Recently, researchers in speech recognition have started to reconsider using whole words as the basic modeling unit, instead of phonetic units. These systems rely on a function that embeds an arbitrary or fixed dimensional speech segments to a vector in a fixed-dimensional space, named acoustic word embedding. Thus, speech segments of words that sound similarly will be projected in a close area in a continuous space. This paper focuses on the evaluation of acoustic word embeddings. We propose two approaches to evaluate the intrinsic performances of acoustic word embeddings in comparison to orthographic representations in order to evaluate whether they capture discriminative phonetic information. Since French language is targeted in experiments, a particular focus is made on homophone words.

No abstract