Atribuição de Descritores a Acórdãos do Supremo Tribunal de Justiça Português com base em Representações Locais Esparsas

• Martim Zanatti ^[1] ; Ricardo Ribeiro ^[2] ; H. Sofia Pinto ^[2] ; José Borbinha ^[2]
  1. [1] Instituto Superior Técnico
  2. [2] INESC-ID, Portugal
• Localización: Linguamática, ISSN 1647-0818, Vol. 17, Nº. 1, 2025, págs. 35-51
• Idioma: portugués
• DOI: 10.21814/lm.17.1.481
• Títulos paralelos:
  • SLEEC assignment of descriptors to judgments of the Supreme Court of Justice of Portugal
• Enlaces
  • Texto completo
• Resumen
  • português

    A Classificação Extrema Multi-etiqueta (XML) consiste na predição de múltiplas etiquetas para um determinado input, sendo um problema fundamental em domínios como categorização de texto, sistemas de recomendação e marcação de imagens. Esta tarefa apresenta desafios significativos para a aprendizagem automática e a recuperação de informação, especialmente devido ao crescimento exponencial de dados online e à consequente necessidade de algoritmos capazes de lidar com conjuntos de dados de grande escala e com um elevado número de etiquetas. Os métodos tradicionais de classificação são inadequados para esta tarefa devido ao vasto número de possíveis combinações de etiquetas e à dispersão das atribuições. Este artigo apresenta os resultados de um projeto realizado com o Supremo Tribunal de Justiça de Portugal, onde abordámos este problema utilizando Sparse Local Embeddings for Extreme Multi-label Classification (SLEEC), uma abordagem baseada em embeddings que demonstrou resultados promissores no domínio legal. O nosso objetivo foi associar descritores, que categorizam os acórdãos do tribunal Português, aos respetivos acórdãos. Este trabalho enfrentou diversos desafios, nos quais se incluem um elevado número de descritores, um conjunto de dados desbalanceado, a presença de muitas etiquetas raras (tail labels) e a extensão considerável dos documentos. Os resultados experimentais demonstram que a nossa abordagem alcançou uma variação de precisão/cobertura entre 0,57 e 0,68, indicando um desempenho promissor nesta tarefa complexa.

  • English

    Extreme Multi-label Classification (XML) involves predicting multiple labels for a given input, a fundamental problem in domains such as text categorization, recommendation systems, and image tagging. This task presents significant challenges for machine learning and information retrieval, particularly given the exponential growth of online data and the concomitant need for algorithms capable of handling large-scale datasets with numerous labels. Traditional classification methods are inadequate for this task due to the vast number of possible label combinations and the sparsity of label assignments. This paper reports the results of a project with the Supreme Court of Justice of Portugal (“Supremo Tribunal de Justi¸ca Portuguˆes”) to address the problem using Sparse Local Embeddings for Extreme Multi-label Classification (SLEEC), an embedding-based approach that showed promising results in legal datasets. Our goal was to associate descriptors, which categorize court judgments, with the judgments themselves. This work tackled various challenges, including a large number of descriptors, an unbalanced dataset, numerous tail labels, and extensive document lengths. Our experimental results demonstrate that our approach achieved a precision/recall variation ranging between 0.57 and 0.68, indicating promising performance in this complex task

• Referencias bibliográficas
  • Agrawal, Rahul, Archit Gupta, Yashoteja Prabhu & Manik Varma. 2013. Multilabel learning with millions of labels: recommending advertiser...
  • Babbar, Rohit & Bernhard Schölkopf. 2017. DiSMEC: Distributed sparse machines for extreme multi-label classification. Em 10th International...
  • Babbar, Rohit & Bernhard Schölkopf. 2018. Adversarial extreme multi-label classification. arXiv [stat.ML/cs.LG]. 10.48550/arXiv.1803.01570
  • Bhatia, Kush, Himanshu Jain, Purushottam Kar, Manik Varma & Prateek Jain. 2015. Sparse local embeddings for extreme multi-label classification....
  • Bi, Wei & James Tin-Yau Kwok. 2013. Efficient multi-label classification with many labels. Em 30th International Conference on Machine Learning...
  • Boutsidis, Christos, Michael W. Mahoney & Petros Drineas. 2009. An improved approximation algorithm for the column subset selection problem....
  • Cissé, Moustapha, Nicolas Usunier, Thierry Arti`eres & Patrick Gallinari. 2013. Robust bloom filters for large multilabel classification tasks....
  • Dong, Wei, Moses Charikar & Kai Li. 2011. Efficient k-nearest neighbor graph construction for generic similarity measures. Em 20th International...
  • Fang, Huang, Minhao Cheng, Cho-Jui Hsieh & Michael P. Friedlander. 2019. Fast training for large-scale one-versus-all linear classifiers...
  • Jain, Himanshu, Yashoteja Prabhu & Manik Varma. 2016. Extreme multi-label loss functions for recommendation, tagging, ranking & other...
  • Jain, Prateek, Raghu Meka & Inderjit S. Dhillon. 2010. Guaranteed rank minimization via singular value projection. Em Advances in Neural Information...
  • Jiang, Ting, Deqing Wang, Leilei Sun, Huayi Yang, Zhengyang Zhao & Fuzhen Zhuang. 2021. LightXML: Transformer with dynamic negative sampling...
  • Kanapala, Ambedkar, Sukomal Pal & Rajendra Pamula. 2019. Text summarization from legal documents: a survey. Artificial Intelligence Review...
  • Khandagale, Sujay, Han Xiao & Rohit Babbar. 2019. Bonsai - diverse and shallow trees for extreme multi-label classification. arXiv [cs.LG/stat.ML]. 10.48550/arXiv.1904.08249
  • Medini, Tharun, Qixuan Huang, Yiqiu Wang, Vijai Mohan & Anshumali Shrivastava. 2019. Extreme classification in log memory using countmin...
  • Melo, Rui, Pedro A. Santos & João Dias. 2023. A semantic search system for the supremo tribunal de justi¸ca. Em 22nd Portuguese Conference...
  • Niculescu-Mizil, Alexandru & Ehsan Abbasnejad. 2017. Label filters for large scale multilabel classification. Em 20th International Conference...
  • Papanikolaou, Yannis, Ioannis Katakis & Grigorios Tsoumakas. 2016. Hierarchical partitioning of the output space in multi-label data....
  • Pennington, Jeffrey, Richard Socher & Christopher Manning. 2014. GloVe: Global vectors for word representation. Em Conference on Empirical...
  • Prabhu, Yashoteja & Manik Varma. 2014. FastXML: a fast, accurate and stable treeclassifier for extreme multi-label learning. Em International...
  • Souza, Fábio, Rodrigo Frassetto Nogueira & Roberto de Alencar Lotufo. 2020. BERTimbau: Pretrained BERT models for Brazilian Portuguese....
  • Tagami, Yukihiro. 2017. AnnexML: Approximate nearest neighbor search for extreme multi-label classification. Em International Conference on...
  • Tai, Farbound & Hsuan-Tien Lin. 2012. Multilabel classification with principal label space transformation. Neural Computation 24(9). 2508–2542....
  • Tsoumakas, Grigorios, Ioannis Katakis & Ioannis Vlahavas. 2008. Effective and efficient multilabel classification in domains with large...
  • Turtle, Howard. 1995. Text retrieval in the legal world. Artificial Intelligence and Law 3(1). 5– 54. 10.1007/BF00877694
  • Wang, Bingyu, Li Chen, Wei Sun, Kechen Qin, Kefeng Li & Hui Zhou. 2019. Ranking-based autoencoder for extreme multilabel classification....
  • Wei, Tong, Zhen Mao, Jiang-Xin Shi, Yu-Feng Li & Min-Ling Zhang. 2022. A survey on extreme multi-label learning. arXiv [cs.LG]. 10.48550/arXiv.2210.03968
  • Yen, Ian En-Hsu, Xiangru Huang, Pradeep Ravikumar, Kai Zhong & Inderjit S. Dhillon. 2016. PD-Sparse : A primal and dual sparse approach...
  • You, Ronghui, Suyang Dai, Zihan Zhang, Hiroshi Mamitsuka & Shanfeng Zhu. 2018. AttentionXML: Extreme multi-label text classification with...
  • Zanatti, Martim, Ricardo Ribeiro & H. Sofia Pinto. 2024. Segmenting model for portuguese judgments. Em Portuguese Conference on Artificial...
  • Zhang, Wenjie, Liwei Wang, Junchi Yan, Xiangfeng Wang & Hongyuan Zha. 2017. Deep extreme multi-label learning. arXiv [cs.LG]. 10.48550/arXiv.1704.03718