Sandra Maria Aluísio

Daniel Peixoto Pinto da Silva, Edresson Casanova, Lucas Rafael Stefanel Gris et al.

During the outbreak of COVID-19 pandemic, several research areas joined efforts to mitigate the damages caused by SARS-CoV-2. In this paper we present an interpretability analysis of a convolutional neural network based model for COVID-19 detection in audios. We investigate which features are important for model decision process, investigating spectrograms, F0, F0 standard deviation, sex and age. Following, we analyse model decisions by generating heat maps for the trained models to capture their attention during the decision process. Focusing on a explainable Inteligence Artificial approach, we show that studied models can taken unbiased decisions even in the presence of spurious data in the training set, given the adequate preprocessing steps. Our best model has 94.44% of accuracy in detection, with results indicating that models favors spectrograms for the decision process, particularly, high energy areas in the spectrogram related to prosodic domains, while F0 also leads to efficient COVID-19 detection.

Edresson Casanova, Christopher Shulby, Alexander Korolev et al.

We explore cross-lingual multi-speaker speech synthesis and cross-lingual voice conversion applied to data augmentation for automatic speech recognition (ASR) systems in low/medium-resource scenarios. Through extensive experiments, we show that our approach permits the application of speech synthesis and voice conversion to improve ASR systems using only one target-language speaker during model training. We also managed to close the gap between ASR models trained with synthesized versus human speech compared to other works that use many speakers. Finally, we show that it is possible to obtain promising ASR training results with our data augmentation method using only a single real speaker in a target language.

Lucas Rafael Stefanel Gris, Arnaldo Candido Junior, Vinícius G. dos Santos et al.

The NURC Project that started in 1969 to study the cultured linguistic urban norm spoken in five Brazilian capitals, was responsible for compiling a large corpus for each capital. The digitized NURC/SP comprises 375 inquiries in 334 hours of recordings taken in São Paulo capital. Although 47 inquiries have transcripts, there was no alignment between the audio-transcription, and 328 inquiries were not transcribed. This article presents an evaluation and error analysis of three automatic speech recognition models trained with spontaneous speech in Portuguese and one model trained with prepared speech. The evaluation allowed us to choose the best model, using WER and CER metrics, in a manually aligned sample of NURC/SP, to automatically transcribe 284 hours.

Rodrigo Lima, Sidney Evaldo Leal, Arnaldo Candido Junior et al.

We present a freely available spontaneous speech corpus for the Brazilian Portuguese language and report preliminary automatic speech recognition (ASR) results, using both the Wav2Vec2-XLSR-53 and Distil-Whisper models fine-tuned and trained on our corpus. The NURC-SP Audio Corpus comprises 401 different speakers (204 females, 197 males) with a total of 239.30 hours of transcribed audio recordings. To the best of our knowledge, this is the first large Paulistano accented spontaneous speech corpus dedicated to the ASR task in Portuguese. We first present the design and development procedures of the NURC-SP Audio Corpus, and then describe four ASR experiments in detail. The experiments demonstrated promising results for the applicability of the corpus for ASR. Specifically, we fine-tuned two versions of Wav2Vec2-XLSR-53 model, trained a Distil-Whisper model using our dataset with labels determined by Whisper Large-V3 model, and fine-tuned this Distil-Whisper model with our corpus. Our best results were the Distil-Whisper fine-tuned over NURC-SP Audio Corpus with a WER of 24.22% followed by a fine-tuned versions of Wav2Vec2-XLSR-53 model with a WER of 33.73%, that is almost 10% point worse than Distil-Whisper's. To enable experiment reproducibility, we share the NURC-SP Audio Corpus dataset, pre-trained models, and training recipes in Hugging-Face and Github repositories.

Arnaldo Candido Junior, Edresson Casanova, Anderson Soares et al.

Automatic Speech recognition (ASR) is a complex and challenging task. In recent years, there have been significant advances in the area. In particular, for the Brazilian Portuguese (BP) language, there were about 376 hours public available for ASR task until the second half of 2020. With the release of new datasets in early 2021, this number increased to 574 hours. The existing resources, however, are composed of audios containing only read and prepared speech. There is a lack of datasets including spontaneous speech, which are essential in different ASR applications. This paper presents CORAA (Corpus of Annotated Audios) v1. with 290.77 hours, a publicly available dataset for ASR in BP containing validated pairs (audio-transcription). CORAA also contains European Portuguese audios (4.69 hours). We also present a public ASR model based on Wav2Vec 2.0 XLSR-53 and fine-tuned over CORAA. Our model achieved a Word Error Rate of 24.18% on CORAA test set and 20.08% on Common Voice test set. When measuring the Character Error Rate, we obtained 11.02% and 6.34% for CORAA and Common Voice, respectively. CORAA corpora were assembled to both improve ASR models in BP with phenomena from spontaneous speech and motivate young researchers to start their studies on ASR for Portuguese. All the corpora are publicly available at https://github.com/nilc-nlp/CORAA under the CC BY-NC-ND 4.0 license.

Lucas Rafael Stefanel Gris, Ricardo Marcacini, Arnaldo Candido Junior et al.

Automatic speech recognition (ASR) systems play a key role in applications involving human-machine interactions. Despite their importance, ASR models for the Portuguese language proposed in the last decade have limitations in relation to the correct identification of punctuation marks in automatic transcriptions, which hinder the use of transcriptions by other systems, models, and even by humans. However, recently Whisper ASR was proposed by OpenAI, a general-purpose speech recognition model that has generated great expectations in dealing with such limitations. This chapter presents the first study on the performance of Whisper for punctuation prediction in the Portuguese language. We present an experimental evaluation considering both theoretical aspects involving pausing points (comma) and complete ideas (exclamation, question, and fullstop), as well as practical aspects involving transcript-based topic modeling - an application dependent on punctuation marks for promising performance. We analyzed experimental results from videos of Museum of the Person, a virtual museum that aims to tell and preserve people's life histories, thus discussing the pros and cons of Whisper in a real-world scenario. Although our experiments indicate that Whisper achieves state-of-the-art results, we conclude that some punctuation marks require improvements, such as exclamation, semicolon and colon.

Sidney Evaldo Leal, Magali Sanches Duran, Carolina Evaristo Scarton et al.

This paper presents and makes publicly available the NILC-Metrix, a computational system comprising 200 metrics proposed in studies on discourse, psycholinguistics, cognitive and computational linguistics, to assess textual complexity in Brazilian Portuguese (BP). These metrics are relevant for descriptive analysis and the creation of computational models and can be used to extract information from various linguistic levels of written and spoken language. The metrics in NILC-Metrix were developed during the last 13 years, starting in 2008 with Coh-Metrix-Port, a tool developed within the scope of the PorSimples project. Coh-Metrix-Port adapted some metrics to BP from the Coh-Metrix tool that computes metrics related to cohesion and coherence of texts in English. After the end of PorSimples in 2010, new metrics were added to the initial 48 metrics of Coh-Metrix-Port. Given the large number of metrics, we present them following an organisation similar to the metrics of Coh-Metrix v3.0 to facilitate comparisons made with metrics in Portuguese and English. In this paper, we illustrate the potential of NILC-Metrix by presenting three applications: (i) a descriptive analysis of the differences between children's film subtitles and texts written for Elementary School I and II (Final Years); (ii) a new predictor of textual complexity for the corpus of original and simplified texts of the PorSimples project; (iii) a complexity prediction model for school grades, using transcripts of children's story narratives told by teenagers. For each application, we evaluate which groups of metrics are more discriminative, showing their contribution for each task.

Edresson Casanova, Christopher Shulby, Eren Gölge et al.

In this paper, we propose SC-GlowTTS: an efficient zero-shot multi-speaker text-to-speech model that improves similarity for speakers unseen during training. We propose a speaker-conditional architecture that explores a flow-based decoder that works in a zero-shot scenario. As text encoders, we explore a dilated residual convolutional-based encoder, gated convolutional-based encoder, and transformer-based encoder. Additionally, we have shown that adjusting a GAN-based vocoder for the spectrograms predicted by the TTS model on the training dataset can significantly improve the similarity and speech quality for new speakers. Our model converges using only 11 speakers, reaching state-of-the-art results for similarity with new speakers, as well as high speech quality.

Edresson Casanova, Arnaldo Candido Junior, Christopher Shulby et al.

Speech provides a natural way for human-computer interaction. In particular, speech synthesis systems are popular in different applications, such as personal assistants, GPS applications, screen readers and accessibility tools. However, not all languages are on the same level when in terms of resources and systems for speech synthesis. This work consists of creating publicly available resources for Brazilian Portuguese in the form of a novel dataset along with deep learning models for end-to-end speech synthesis. Such dataset has 10.5 hours from a single speaker, from which a Tacotron 2 model with the RTISI-LA vocoder presented the best performance, achieving a 4.03 MOS value. The obtained results are comparable to related works covering English language and the state-of-the-art in Portuguese.

Edresson Casanova, Arnaldo Candido Junior, Christopher Shulby et al.

In this paper we present an efficient method for training models for speaker recognition using small or under-resourced datasets. This method requires less data than other SOTA (State-Of-The-Art) methods, e.g. the Angular Prototypical and GE2E loss functions, while achieving similar results to those methods. This is done using the knowledge of the reconstruction of a phoneme in the speaker's voice. For this purpose, a new dataset was built, composed of 40 male speakers, who read sentences in Portuguese, totaling approximately 3h. We compare the three best architectures trained using our method to select the best one, which is the one with a shallow architecture. Then, we compared this model with the SOTA method for the speaker recognition task: the Fast ResNet-34 trained with approximately 2,000 hours, using the loss functions Angular Prototypical and GE2E. Three experiments were carried out with datasets in different languages. Among these three experiments, our model achieved the second best result in two experiments and the best result in one of them. This highlights the importance of our method, which proved to be a great competitor to SOTA speaker recognition models, with 500x less data and a simpler approach.

Nathan Hartmann, Erick Fonseca, Christopher Shulby et al.

Word embeddings have been found to provide meaningful representations for words in an efficient way; therefore, they have become common in Natural Language Processing sys- tems. In this paper, we evaluated different word embedding models trained on a large Portuguese corpus, including both Brazilian and European variants. We trained 31 word embedding models using FastText, GloVe, Wang2Vec and Word2Vec. We evaluated them intrinsically on syntactic and semantic analogies and extrinsically on POS tagging and sentence semantic similarity tasks. The obtained results suggest that word analogies are not appropriate for word embedding evaluation; task-specific evaluations appear to be a better option.

Marcos V. Treviso, Christopher D. Shulby, Sandra M. Aluisio

This paper is motivated by the automation of neuropsychological tests involving discourse analysis in the retellings of narratives by patients with potential cognitive impairment. In this scenario the task of sentence boundary detection in speech transcripts is important as discourse analysis involves the application of Natural Language Processing tools, such as taggers and parsers, which depend on the sentence as a processing unit. Our aim in this paper is to verify which embedding induction method works best for the sentence boundary detection task, specifically whether it be those which were proposed to capture semantic, syntactic or morphological similarities.

Christopher Dane Shulby, Martha Dais Ferreira, Rodrigo F. de Mello et al.

High-accuracy speech recognition is especially challenging when large datasets are not available. It is possible to bridge this gap with careful and knowledge-driven parsing combined with the biologically inspired CNN and the learning guarantees of the Vapnik Chervonenkis (VC) theory. This work presents a Shallow-CNN-HTSVM (Hierarchical Tree Support Vector Machine classifier) architecture which uses a predefined knowledge-based set of rules with statistical machine learning techniques. Here we show that gross errors present even in state-of-the-art systems can be avoided and that an accurate acoustic model can be built in a hierarchical fashion. The CNN-HTSVM acoustic model outperforms traditional GMM-HMM models and the HTSVM structure outperforms a MLP multi-class classifier. More importantly we isolate the performance of the acoustic model and provide results on both the frame and phoneme level considering the true robustness of the model. We show that even with a small amount of data accurate and robust recognition rates can be obtained.

Leandro B. dos Santos, Magali S. Duran, Nathan S. Hartmann et al.

Psycholinguistic properties of words have been used in various approaches to Natural Language Processing tasks, such as text simplification and readability assessment. Most of these properties are subjective, involving costly and time-consuming surveys to be gathered. Recent approaches use the limited datasets of psycholinguistic properties to extend them automatically to large lexicons. However, some of the resources used by such approaches are not available to most languages. This study presents a method to infer psycholinguistic properties for Brazilian Portuguese (BP) using regressors built with a light set of features usually available for less resourced languages: word length, frequency lists, lexical databases composed of school dictionaries and word embedding models. The correlations between the properties inferred are close to those obtained by related works. The resulting resource contains 26,874 words in BP annotated with concreteness, age of acquisition, imageability and subjective frequency.

Leandro B. dos Santos, Edilson A. Corrêa, Osvaldo N. Oliveira et al.

Mild Cognitive Impairment (MCI) is a mental disorder difficult to diagnose. Linguistic features, mainly from parsers, have been used to detect MCI, but this is not suitable for large-scale assessments. MCI disfluencies produce non-grammatical speech that requires manual or high precision automatic correction of transcripts. In this paper, we modeled transcripts into complex networks and enriched them with word embedding (CNE) to better represent short texts produced in neuropsychological assessments. The network measurements were applied with well-known classifiers to automatically identify MCI in transcripts, in a binary classification task. A comparison was made with the performance of traditional approaches using Bag of Words (BoW) and linguistic features for three datasets: DementiaBank in English, and Cinderella and Arizona-Battery in Portuguese. Overall, CNE provided higher accuracy than using only complex networks, while Support Vector Machine was superior to other classifiers. CNE provided the highest accuracies for DementiaBank and Cinderella, but BoW was more efficient for the Arizona-Battery dataset probably owing to its short narratives. The approach using linguistic features yielded higher accuracy if the transcriptions of the Cinderella dataset were manually revised. Taken together, the results indicate that complex networks enriched with embedding is promising for detecting MCI in large-scale assessments

Nathan Siegle Hartmann, Magali Sanches Duran, Sandra Maria Aluísio

Semantic Role Labeling (SRL) is a Natural Language Processing task that enables the detection of events described in sentences and the participants of these events. For Brazilian Portuguese (BP), there are two studies recently concluded that perform SRL in journalistic texts. [1] obtained F1-measure scores of 79.6, using the PropBank.Br corpus, which has syntactic trees manually revised, [8], without using a treebank for training, obtained F1-measure scores of 68.0 for the same corpus. However, the use of manually revised syntactic trees for this task does not represent a real scenario of application. The goal of this paper is to evaluate the performance of SRL on revised and non-revised syntactic trees using a larger and balanced corpus of BP journalistic texts. First, we have shown that [1]'s system also performs better than [8]'s system on the larger corpus. Second, the SRL system trained on non-revised syntactic trees performs better over non-revised trees than a system trained on gold-standard data.

Nathan Siegle Hartmann, Livia Cucatto, Danielle Brants et al.

Recent research shows that most Brazilian students have serious problems regarding their reading skills. The full development of this skill is key for the academic and professional future of every citizen. Tools for classifying the complexity of reading materials for children aim to improve the quality of the model of teaching reading and text comprehension. For English, Fengs work [11] is considered the state-of-art in grade level prediction and achieved 74% of accuracy in automatically classifying 4 levels of textual complexity for close school grades. There are no classifiers for nonfiction texts for close grades in Portuguese. In this article, we propose a scheme for manual annotation of texts in 5 grade levels, which will be used for customized reading to avoid the lack of interest by students who are more advanced in reading and the blocking of those that still need to make further progress. We obtained 52% of accuracy in classifying texts into 5 levels and 74% in 3 levels. The results prove to be promising when compared to the state-of-art work.9

Marcos Vinícius Treviso, Christopher Shulby, Sandra Maria Aluísio

Automated discourse analysis tools based on Natural Language Processing (NLP) aiming at the diagnosis of language-impairing dementias generally extract several textual metrics of narrative transcripts. However, the absence of sentence boundary segmentation in the transcripts prevents the direct application of NLP methods which rely on these marks to function properly, such as taggers and parsers. We present the first steps taken towards automatic neuropsychological evaluation based on narrative discourse analysis, presenting a new automatic sentence segmentation method for impaired speech. Our model uses recurrent convolutional neural networks with prosodic, Part of Speech (PoS) features, and word embeddings. It was evaluated intrinsically on impaired, spontaneous speech, as well as, normal, prepared speech, and presents better results for healthy elderly (CTL) (F1 = 0.74) and Mild Cognitive Impairment (MCI) patients (F1 = 0.70) than the Conditional Random Fields method (F1 = 0.55 and 0.53, respectively) used in the same context of our study. The results suggest that our model is robust for impaired speech and can be used in automated discourse analysis tools to differentiate narratives produced by MCI and CTL.