|
EPIA'03 - 11th Portuguese Conference on Artificial Intelligence
NLTR -- Natural Language and Text Retrieval
|
Session: December 5, 10:45-12:15, Room B |
Title: |
A preliminary approach to the multilabel classification problem of Portuguese juridical documents |
|
Teresa Gonçalves and Paulo Quaresma |
Abstract: |
Portuguese juridical documents from Supreme Courts and
the Attorney General's Office are manually classified by juridical experts into a set of classes belonging to a taxonomy of concepts.
In this paper a preliminary approach to develop techniques to automatically classify these juridical documents is proposed.
As basic strategy, the integration of natural language processing techniques with machine learning ones is used.
Support Vector Machines (SVM) are used as learning algorithm
and the obtained results are presented and compared with other approaches, such as, C4.5, and naive Bayes. |
Back to schedule. |