Метод нейромережевого виявлення кібербулінгу з використанням хмарних сервісів та об'єктно-орієнтованої моделі
Вантажиться...
Дата
2024
Назва журналу
Номер ISSN
Назва тому
Видавець
Хмельницький національний університет
Анотація
У роботі пропонується практичний підхід до виявлення кібербулінгу із використанням нейронної мережі BiLSTM, навченої за допомогою хмарних сервісів та застосунку, що реалізовує запропонований метод виявлення кібербулінгу. Для дослідження ефективності запропонованого методу було створено об’єктно-орієнтовану програмну реалізацію середовища програмування PyCharm, а також ноутбук для виконання в хмарному сервісі «Google Colab» для навчання нейромережі.
The article proposes the method for detecting cyberbullying in posts of social Internet networks, designed for the automated detection of cyberbullying in text messages published in social networks using a neural network approach. The current state of the field of automated detection of cyberbullying is considered, where, based on the studied material, it is proposed to create a method of neural network detection of cyberbullying using cloud services and an object-oriented model. To investigate the effectiveness of the proposed method, a software implementation written in the Python programming language in the PyCharm programming environment was created, as well as a laptop for execution in the Google Colab cloud service for neural network training. The English-language "Cyberbullying Classification" dataset consisting of 39,747 samples was used as research data, from which 8,000 samples containing cyberbullying and 7,945 samples without cyberbullying were selected. The proposed approach based on the object-oriented model contributes to the creation of flexible, adaptive and easily expandable systems, and the use of cloud services for training neural networks provides an opportunity not to be limited exclusively to the resources of a physical personal computer, but to use specialized computing resources, such as graphics processors (GPU) or tensor processors (TPU), which significantly accelerate the process of training neural networks. The approach has some limitations, the main limitation of the proposed approach is the maximum length of the input sequence, which is 500 characters. Considering the average length of a tweet is about 200 characters, such a limitation is reasonable. In order to improve the result, you can supplement the training dataset and specify some restrictions on the minimum and maximum length of records.
The article proposes the method for detecting cyberbullying in posts of social Internet networks, designed for the automated detection of cyberbullying in text messages published in social networks using a neural network approach. The current state of the field of automated detection of cyberbullying is considered, where, based on the studied material, it is proposed to create a method of neural network detection of cyberbullying using cloud services and an object-oriented model. To investigate the effectiveness of the proposed method, a software implementation written in the Python programming language in the PyCharm programming environment was created, as well as a laptop for execution in the Google Colab cloud service for neural network training. The English-language "Cyberbullying Classification" dataset consisting of 39,747 samples was used as research data, from which 8,000 samples containing cyberbullying and 7,945 samples without cyberbullying were selected. The proposed approach based on the object-oriented model contributes to the creation of flexible, adaptive and easily expandable systems, and the use of cloud services for training neural networks provides an opportunity not to be limited exclusively to the resources of a physical personal computer, but to use specialized computing resources, such as graphics processors (GPU) or tensor processors (TPU), which significantly accelerate the process of training neural networks. The approach has some limitations, the main limitation of the proposed approach is the maximum length of the input sequence, which is 500 characters. Considering the average length of a tweet is about 200 characters, such a limitation is reasonable. In order to improve the result, you can supplement the training dataset and specify some restrictions on the minimum and maximum length of records.
Опис
Ключові слова
виявлення кібербулінгу, нейронна мережа, об'єктно-орієнтована модель, хмарні сервіси, Google Colab, BiLSTM, твіт, cyberbullying detection, neural network, object-oriented model, cyberbullying, cloud services, tweet
Бібліографічний опис
Метод нейромережевого виявлення кібербулінгу з використанням хмарних сервісів та об'єктно-орієнтованої моделі / М. Молчанова, О. Мазурець, О. Собко, В. Кліменко, В. Андрощук // Вісник Хмельницького національного університету. Технічні науки. – 2024. – № 2. – С. 200-206.