Text

KLASIFIKASI UJARAN KEBENCIAN MENGGUNAKAN ARSITEKTUR TRANSFORMER, CNN, DAN GRU

---

Penilaian

---

Tweets containing hate speech sometimes have several categories. Hate speech in this study has 12 label categories, namely hate speech, abusive words, hate speech against individuals, hate speech against groups, hate speech against religion, hate speech against ethnicity or race, hate speech against physical conditions, hate speech against gender, other hate speech, weak hate speech, medium hate speech, and strong hate speech. Automatic classification can help early detection of hate speech. This research proposes a combination of Transformer, CNN, and GRU architecture. The transformer is placed in the first block to obtain global information from the text. CNN is placed after the transformer block to extract important information from the transformer. GRU is placed in the last block to help retain and remove information efficiently from the data generated by CNN. Performance evaluation is done by measuring accuracy, precision, recall, and f1-score. Accuracy is 96.39%, indicating that the model is able to predict hate speech correctly as a whole. Precision is 97.3%, which means the model can predict hate speech well. Recall is 97.77%, indicating that the model is able to predict most of the hate speech well. F1-Score is 97.53% which shows a good balance between precision and recall in predicting hate speech. The high f1-score result reflects the stable performance of the model in detecting and minimizing prediction errors. The performance evaluation results per label show that the hate speech label against gender has the highest accuracy, recall, and f1-score results with 99.24%, 99.69%, and 99.61% respectively, while the hate speech label against ethnicity or race has the highest precision result of 99.64%. On hate speech labels, the precision and f1-score results are still below 95%. This research shows that the use of a combination model of Transformer, CNN, and GRU can increase the effectiveness of hate speech classification in Indonesian.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2507001952	T169921	T1699212025	Central Library (Reference)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T1699212025

Publisher

: Prodi Ilmu Matematika, Fakultas Matematika Dan Ilmu Pengetahuan Alam Universitas Sriwijaya., 2025

Collation

xviii, 67 hlm.: Ilus., tab. 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

510.07

Content Type

-

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Matematika

Specific Detail Info

-

Statement of Responsibility

EM

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment