Data Augmentation Using Transformers and Similarity Measures for Improving Arabic Text Classification
The performance of learning models heavily relies on the availability and adequacy of training data. To address the dataset adequacy issue, researchers have extensively explored data augmentation (DA) as a promising approach. DA generates new data instances through transformations applied to the ava...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10328600/ |
| _version_ | 1797321120331857920 |
|---|---|
| author | Dania Refai Saleh Abu-Soud Mohammad J. Abdel-Rahman |
| author_facet | Dania Refai Saleh Abu-Soud Mohammad J. Abdel-Rahman |
| author_sort | Dania Refai |
| collection | DOAJ |
| description | The performance of learning models heavily relies on the availability and adequacy of training data. To address the dataset adequacy issue, researchers have extensively explored data augmentation (DA) as a promising approach. DA generates new data instances through transformations applied to the available data, thereby increasing dataset size and variability. This approach has enhanced model performance and accuracy, particularly in addressing class imbalance problems in classification tasks. However, few studies have explored DA for the Arabic language, relying on traditional approaches such as paraphrasing or noising-based techniques. In this paper, we propose a new Arabic DA method that employs the recent powerful modeling technique, namely the AraGPT-2, for the augmentation process. The generated sentences are evaluated in terms of context, semantics, diversity, and novelty using the Euclidean, cosine, Jaccard, and BLEU distances. Finally, the AraBERT transformer is used on sentiment classification tasks to evaluate the classification performance of the augmented Arabic dataset. The experiments were conducted on four sentiment Arabic datasets: AraSarcasm, ASTD, ATT, and MOVIE. The selected datasets vary in size, label number, and unbalanced classes. The results show that the proposed methodology enhanced the Arabic sentiment text classification on all datasets with an increase in F1 score by 7% in AraSarcasm, 8% in ASTD, 11% in ATT, and 13% in MOVIE. |
| first_indexed | 2024-03-08T04:52:57Z |
| format | Article |
| id | doaj.art-d4b36ed458034b7aa1767aa4e34a39e9 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-03-08T04:52:57Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-d4b36ed458034b7aa1767aa4e34a39e92024-02-08T00:00:36ZengIEEEIEEE Access2169-35362023-01-011113251613253110.1109/ACCESS.2023.333631110328600Data Augmentation Using Transformers and Similarity Measures for Improving Arabic Text ClassificationDania Refai0https://orcid.org/0000-0002-4599-8797Saleh Abu-Soud1https://orcid.org/0000-0001-9144-9409Mohammad J. Abdel-Rahman2https://orcid.org/0000-0001-5788-6656Computer Science Department, Princess Sumaya University for Technology, Amman, JordanData Science Department, Princess Sumaya University for Technology, Amman, JordanData Science Department, Princess Sumaya University for Technology, Amman, JordanThe performance of learning models heavily relies on the availability and adequacy of training data. To address the dataset adequacy issue, researchers have extensively explored data augmentation (DA) as a promising approach. DA generates new data instances through transformations applied to the available data, thereby increasing dataset size and variability. This approach has enhanced model performance and accuracy, particularly in addressing class imbalance problems in classification tasks. However, few studies have explored DA for the Arabic language, relying on traditional approaches such as paraphrasing or noising-based techniques. In this paper, we propose a new Arabic DA method that employs the recent powerful modeling technique, namely the AraGPT-2, for the augmentation process. The generated sentences are evaluated in terms of context, semantics, diversity, and novelty using the Euclidean, cosine, Jaccard, and BLEU distances. Finally, the AraBERT transformer is used on sentiment classification tasks to evaluate the classification performance of the augmented Arabic dataset. The experiments were conducted on four sentiment Arabic datasets: AraSarcasm, ASTD, ATT, and MOVIE. The selected datasets vary in size, label number, and unbalanced classes. The results show that the proposed methodology enhanced the Arabic sentiment text classification on all datasets with an increase in F1 score by 7% in AraSarcasm, 8% in ASTD, 11% in ATT, and 13% in MOVIE.https://ieeexplore.ieee.org/document/10328600/ArabicAraBERTAraGPT-2data augmentationmachine learningnatural language processing |
| spellingShingle | Dania Refai Saleh Abu-Soud Mohammad J. Abdel-Rahman Data Augmentation Using Transformers and Similarity Measures for Improving Arabic Text Classification IEEE Access Arabic AraBERT AraGPT-2 data augmentation machine learning natural language processing |
| title | Data Augmentation Using Transformers and Similarity Measures for Improving Arabic Text Classification |
| title_full | Data Augmentation Using Transformers and Similarity Measures for Improving Arabic Text Classification |
| title_fullStr | Data Augmentation Using Transformers and Similarity Measures for Improving Arabic Text Classification |
| title_full_unstemmed | Data Augmentation Using Transformers and Similarity Measures for Improving Arabic Text Classification |
| title_short | Data Augmentation Using Transformers and Similarity Measures for Improving Arabic Text Classification |
| title_sort | data augmentation using transformers and similarity measures for improving arabic text classification |
| topic | Arabic AraBERT AraGPT-2 data augmentation machine learning natural language processing |
| url | https://ieeexplore.ieee.org/document/10328600/ |
| work_keys_str_mv | AT daniarefai dataaugmentationusingtransformersandsimilaritymeasuresforimprovingarabictextclassification AT salehabusoud dataaugmentationusingtransformersandsimilaritymeasuresforimprovingarabictextclassification AT mohammadjabdelrahman dataaugmentationusingtransformersandsimilaritymeasuresforimprovingarabictextclassification |