Brazil

Apresentamos a instanciação que realizamos das diretrizes do modelo Universal Dependencies (UD) para a anotação de Numerais em português. Apresentamos as diferenças do conceito da classe de numerais na Nomenclatura Gramatical Brasileira e Portuguesa e nas diretrizes da UD. Apresentamos os principais problemas detectados . Os resultados da combinação das diretrizes da UD com as características dos numerais em português são apresentados em detalhes, com exemplos, juntamente com os argumentos que amparam cada decisão tomada. 


BRACIS 2021

Descrição de numerais segundo modelo Universal Dependencies e sua anotação no português

numeral annotation

portuguese language

universal dependencies

**BRACIS** originated from the combination of the two most important scientific events in Brazil in Artificial Intelligence (AI) and Computational Intelligence (CI): the Brazilian Symposium on Artificial Intelligence – **SBIA** (22 editions), and the Brazilian Symposium on Neural Networks – **SBRN** (13 editions). **ENIAC** is a national forum for researchers, practitioners, educators and students to present and discuss innovations, trends, experiences, developments and work in progress in the fields of Artificial and Computational Intelligence. **STIL** is the bi-annual Language Technology event.
All events are supported by the Brazilian Computer Society – Special Interest Groups: **CE-IA**, **CE-IC** and **CE-PLN**.

To access the 10th Brazilian Conference on Intelligent Systems (**BRACIS**), you need to register for the conference.  Please visit the Conference official site to register

[Register now](http://c4ai.inova.usp.br/bracis2021/)

Register now

BRACIS originated from the combination of the two most important scientific events in Brazil in Artificial Intelligence (AI) and Computational Intelligence (CI): the Brazilian Symposium on Artificial Intelligence – SBIA (22 editions), and the Brazilian Symposium on Neural Networks – SBRN (13 editions). 
ENIAC is a national forum for researchers, practitioners, educators and students to present and discuss innovations, trends, experiences, developments and work in progress in the fields of Artificial and Computational Intelligence. 
STIL is the bi-annual Language Technology event.
All events are supported by the Brazilian Computer Society – Special Interest Groups: CE-IA, CE-IC and CE-PLN

technical paper

Apresentamos nosso trabalho no qual realizamos o fine-tuning dos modelos BERTimbau-base e BERTimbau-large na tarefa de classificação de 27 tipos de emoções em sentenças, baseado no dataset GoEmotions traduzido para a língua portuguesa, por meio de ferramentas de traducão automática. Comparamos os resultados de nossos experimentos com os resultados disponibilizados pelos autores do dataset GoEmotions e obtivemos um ganho de desempenho ao qual atribuímos ao algoritmo de balanceamento utilizado. Além disso avaliamos o modelo BERTimbau em um dataset de tweets em português.

Utilizando BERTimbau para a Classificação de Emoções em Português

This study introduces novel data and models for the task of Sentiment
Analysis in Portuguese texts about Diabetes Mellitus. The corpus contains
1290 posts retrieved from online health community forums in Portuguese and
annotated by two annotators according to 3 sentiment categories (e.g. Positive,
Neutral and Negative). Evaluation of traditional (Support Vector Machine,
Decision Tree, Random Forest and Logistic Regression classifiers) and state-ofthe-
art (BERT-based models) machine learning classifiers for the task showed
the advantage in performance of the latter models as expected. Data and models
are available to the community upon request.

Sentiment Analysis in Portuguese Texts from Online Health Community Forums: Data, Model and Evaluation

Affective Computing is the study of how computers can recognize, interpret and simulate human affects. Sentiment Analysis is a common task inNLP related to this topic, but it focuses only on emotion valence (positive, negative, neutral). An emerging approach in NLP is Emotion Recognition, which relies on fined-grained classification. This research describes an approach to create a lexical-based weakly supervised corpus for fine-grained emotion in Portuguese. We evaluated our dataset by fine-tuning a transformer-based language model (BERT) and validating it on a Gold Standard annotated validation set. Our results (F1-score=.64) suggest lexical-based weak supervision as an appropriate strategy for initial work in low resourced environment.

A Weakly Supervised Dataset of Fine-Grained Emotions in Portuguese

Bullying and cyberbullying are words commonly seen in today's news. Although the scientific community has evaluated text mining techniques for cyberbullying detection, few studies have targeted Brazilian Portuguese datasets. Our study aims to assess the text mining application to detect cyberbullying messages written in Brazilian Portuguese. We gathered posts and comments from Reddit communities and extracted several text features. We then processed these features using Naïve Bayes and SVM classifiers to uncover cyberbullying activity. The outcomes of this experiment may not be used solo for cyberbullying detection; however, they can aid moderators in prioritizing content reviews and acting faster on real cyberbullying cases.

Text Mining for Cyberbullying Detection: a Brazilian Portuguese Evaluation

This study presents a new version of the lexical decision test designed to capture the social appreciation of variable phonological phenomena [Freitag and Souza 2019]. The effects of lexical decision were tested in standard and non-standard variants of the phenomena of monophthongtion descending and increasing, denasalization of final unstressed nasal diphthong and palatalization of alveolar stops, in progressive and regressive environment, in a sample of 25 university students from Sergipe. The results reinforce what has already been identified in the previous study: stereotype and marker type variants are significantly associated with non-word; while the indicator variants do not show this relation.

Efeitos da variação linguística na decisão lexical

Apresentamos uma análise comparativa acerca da palatalização de
/t/ e /d/ em duas amostras independentes, e com tamanho distinto, compostas por dados de fala (n = 8.850) e de leitura em voz alta (n = 831) de estudantes da Universidade Federal de Sergipe. Controlamos as variáveis sociais deslocamento e tempo no curso, e as variáveis linguísticas contexto anterior, contexto posterior, tonicidade e sonoridade, com o objetivo de identificar se os efeitos condicionantes da palatalização da fala atuam na leitura em voz alta. Os resultados sinalizam que, mesmo em amostras distintas, o comportamento dos fatores sociais e linguísticos apresentam estabilidade na mudança e propagação da palatalização na comunidade.


Palatalização na fala e na leitura de universitários sergipanos

This work investigates the syntactic-semantic proprieties of the verb “falar” (to speak) of Brazilian Portuguese (BP) in a journalistic corpus and three BP verbal databases (VerbNet.Br, VerboWeb and Verbo-Brasil). The data demonstrate a polyvalence and complexity of “falar”. In addition, concerning the analysis of the databases, it was verified a need for manual reviews, made out by linguists, and/or expansions of the linguistic descriptions, especially when considering the relevance of constructions with the verb "falar" for different applications in the Natural Language Processing area

A propósito do verbo falar no português brasileiro: uma análise em corpus e em bases de dados verbais

Os provérbios são um tipo especial de unidades linguísticas que tem
sido amplamente ignorado pela comunidade de Processamento de Linguagem
Natural (PLN), apesar de levantarem desafios interessantes para o processamento.
Este artigo apresenta o procedimento de integração do Mínimo Paremiológico do Português no sistema STRING e a distribuição desses provérbios
mais usuais em três corpora distintos do português (europeu).

Provérbios portugueses usuais: distribuição em corpora

The annotation of informal texts within the Universal Dependencies framework requires two segmentation processes: definition of the relevant unity for syntactic analysis and identification of syntactic words. In this paper, we present the linguistic idiosyncrasies of DANTEStocks, a corpus of tweets from the financial market, written in Portuguese, and the general guidelines for their automatic segmentation. As such, this work contributes to a better understanding of linguistic aspects of tweets and the development of resources and tools for automatic processing of this subgenre of user-generated content.

Descrição Preliminar do Corpus DANTEStocks: Diretrizes de Segmentação para Anotação segundo Universal Dependencies

This sentiment analysis work is focused on the task of identifying aspects, emphasizing the so-called implicit aspects, i.e., those that are not explicitly mentioned in the texts. For this, we analyzed frequency-based methods, adapted rules from the English language to Portuguese, and developed a method that learns new rules through corpus analysis.

Premium content

Descrição de numerais segundo modelo Universal Dependencies e sua anotação no português

Downloads

Next from BRACIS 2021

Utilizando BERTimbau para a Classificação de Emoções em Português

Similar lecture

Universal Dependencies-based PoS Tagging Refinement through Linguistic Resources