Arabic named entity recognition based on tree- based pipeline optimization tool

BRAHIM AIT BEN ALI; SOUKAINA MIHI; ISMAIL EL BAZI; NABIL LAACHFOUBI

No Detail Found!

Arabic named entity recognition based on tree- based pipeline optimization tool

Author(s)

BRAHIM AIT BEN ALI Hassan First University of Settat, Faculty of Sciences and Techniques, IR2M Laboratory, Settat, Morocco

SOUKAINA MIHI Hassan First University of Settat, Faculty of Sciences and Techniques, IR2M Laboratory, Settat, Morocco

ISMAIL EL BAZI Sultan Moulay Slimane University, National School of Business and Management, Beni Mellal, Morocco

NABIL LAACHFOUBI Hassan First University of Settat, Faculty of Sciences and Techniques, IR2M Laboratory, Settat, Morocco

Abstract

Named Entity Recognition (NER) is a clue task to improve automatic text processing, which is needed in a diverse variety of applications. NER techniques vary from hand-craft rules to machine learning approaches. As a human opting for a supervised learning algorithm, it is often difficult to choose the optimal machine learning (ML) model for a classification problem. Automated ML (AutoML) aims at automatically selecting, composing, and parameterizing machine learning algorithms in order to obtain optimal performance for a given task (dataset). In this paper, we apply an approach based on Tree-based Pipeline Optimization Tool (TPOT). This method uses genetic programming based on the tree structure to find the model and its hyperparameters that more closely predicts the class of Arabic named entities in the text comes from social media. The structure and parameters are fine-tuned to achieve the optimum performance of the machine's learning pipeline. Our method outperforms strong baselines. It reaches a novel state-ofthe-art in the NER task.

Publication Details

Page(s) 2963-2976

DOI DOI not available

Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 98, Issue: 15, Year: 2020

Request the full text from PASTIC

Keywords

Automl, Tree-Based Pipeline Optimization Tool (TPOT)

Citations

Downloads

Views