Impact of Text Specificity and Size on Word Embeddings Performance: An Empirical Evaluation in Brazilian Legal Domain

Pont, Thiago Raulino Dal; Sabo, Isabela Cristina; Hübner, Jomi Fred; Rover, Aires José

doi:10.1007/978-3-030-61377-8_36

Cited by 6 publications

(5 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…For example, "In these processes, the consumer claims for compensation for material or moral damages against an airline company due to failures in its services." [Dal Pont et al 2020]. There is an occurrence of the search string (moral).…”

Section: Methodology and Research Methodsmentioning

confidence: 99%

Ethics: What is the Research Scenario in the Brazilian Conference BRACIS?

Carvalho

Murakami

Suzano

et al. 2022

Anais Do XIX Encontro Nacional De Inteligência Artificial E Computacional (ENIAC 2022)

View full text Add to dashboard Cite

Artificial Intelligence (AI) presents many ethical dilemmas, such as explainability, bias, military uses, surveillance capitalism, employment, and jobs. In the scientific context, AI can lead us to a crisis of reproducibility spread across several areas of knowledge and guide mathematicians to solve high complexity problems. Both companies and government forward their guidelines, recommendations, and materials combining Ethics and AI. In this paper, we investigate the involvement of the Brazilian academic-scientific community with moral or ethical aspects through its publications, covering the Brazilian Conference on Intelligent Systems (BRACIS) as the most prominent Brazilian AI conference. Through a Literature Systematic Review method, we answer the main research question: what is the panorama of the explicit occurrence of ethical aspects in the BRACIS, ENIAC, and STIL conference papers? The results indicate a low occurrence of ethical aspects and increasing behavior over the years. Ethical deliberation was fruitful, constructive, and critical among these few occurrences. Whether in the Brazilian or international context, there are spaces to be filled and open opportunities for exploration along this path.

show abstract

Section: Methodology and Research Methodsmentioning

confidence: 99%

Ethics: What is the Research Scenario in the Brazilian Conference BRACIS?

Carvalho

Murakami

Suzano

et al. 2022

Anais Do XIX Encontro Nacional De Inteligência Artificial E Computacional (ENIAC 2022)

View full text Add to dashboard Cite

show abstract

“…For the text representation, we use Bag of Words using term frequency (TF) values. We also tested word embeddings trained with legal documents written in Portuguese ( Dal Pont et al, 2020 ), MultiLingual Bidirectional Encoder Representations from Transformer (BERT) ( Devlin et al, 2018 ), and TF combined with Inverse Document Frequency (IDF), although TF achieved the best results for the experiments presented in this work. As an example, Multilingual BERT achieved a MAE of 6,192, a RMSE of 7,091, and a of .…”

Section: Proposed Pipeline and Experimentsmentioning

confidence: 99%

Regression applied to legal judgments to predict compensation for immaterial damage

Pont

Sabo

Hübner

et al. 2023

PeerJ Computer Science

View full text Add to dashboard Cite

Immaterial damage compensation is a controversial matter in the judicial practice of several law systems. Due to a lack of criteria for its assessment, the judge is free to establish the value based on his/her conviction. Our research motivation is that knowing the estimated amount of immaterial damage compensation at the initial stage of a lawsuit can encourage an agreement between the parties. We thus investigate text regression techniques to predict the compensation value from legal judgments in which consumers had problems with airlines and claim for immaterial damage. We start from a simple pipeline and create others by adding some natural language processing (NLP) and machine learning (ML) techniques, which we call adjustments. The adjustments include N-Grams Extraction, Feature Selection, Overfitting Avoidance, Cross-Validation and Outliers Removal. An special adjustment, Addition of Attributes Extracted by the Legal Expert (AELE), is proposed as a complementary input to the case text. We evaluate the impact of adding these adjustments in the pipeline in terms of prediction quality and execution time. N-Grams Extraction and Addition of AELE have the biggest impact on the prediction quality. In terms of execution time, Feature Selection and Overfitting Avoidance have significant importance. Moreover, we notice the existence of pipelines with subsets of adjustments that achieved better prediction quality than a pipeline with them all. The result is promising since the prediction error of the best pipeline is acceptable in the legal environment. Consequently, the predictions will likely be helpful in a legal environment.

show abstract

“…Também observando vetores embedings resultantes, [Dal Pont et al 2020] avaliaram o impacto da especificidade e do tamanho do corpus de texto utilizado no treinamento dos vetores. Aplicados a dados jurídicos brasileiros em vários níveis de segmentac ¸ão, os resultados mostraram que corpus menores capturam melhor as especificidades textos.…”

Section: Embeddings Orientado Ao Segmento Jurídicounclassified

Embeddings Jurídico: Representações Orientadas à Linguagem Jurídica Brasileira

Carmo,

Serejo,

Jacob Junior

et al. 2023

Anais Do XI Workshop De Computação Aplicada Em Governo Eletrônico (WCGE 2023)

View full text Add to dashboard Cite

O processamento automático de textos jurídicos dispostos em linguagem natural proporciona o desenvolvimento de diversas aplicações para o setor, como a classificação de processos por assunto, sumarização de documentos, tradução para linguagem cidadã etc. Nesse sentido, o judiciário brasileiro lançou o programa Justiça 4.0, buscando soluções que ofereçam celeridade nas atividades processuais. Convém pontuar que a linguagem técnica predomina nesse domínio de aplicação, o que adiciona desafios para modelagem dos dados, exigindo modelos especializados para o segmento. Frente ao exposto, esse trabalho tem como objetivo a construção de modelos embeddings orientados ao âmbito jurídico visando alimentar aplicações na área. Para isso, foram extraídos aproximadamente 500.000 documentos de instituições de justiça do Brasil das mais variadas esferas (civil, criminal, trabalhista etc). Os modelos foram avaliados por meio da classificação de petições iniciais e os resultados mostraram-se competitivos quando comparados a modelos generalistas da língua portuguesa. Tais resultados mostram que modelos treinados com documentos jurídicos compreendem melhor as especificidades da linguagem do segmento e têm o potencial de fomentar novas aplicações para o setor.

show abstract

Impact of Text Specificity and Size on Word Embeddings Performance: An Empirical Evaluation in Brazilian Legal Domain

Cited by 6 publications

References 19 publications

Ethics: What is the Research Scenario in the Brazilian Conference BRACIS?

Ethics: What is the Research Scenario in the Brazilian Conference BRACIS?

Regression applied to legal judgments to predict compensation for immaterial damage

Embeddings Jurídico: Representações Orientadas à Linguagem Jurídica Brasileira

Contact Info

Product

Resources

About