Scholarlink

<p>The use of chatbots in the fields of business and education has greatly expanded in recent years due to their capacity to automate and facilitate communication with users. The creation of chatbots is a difficult but essential task in natural language processing (NLP). Deep learning (DL) approaches have advanced significantly in recent years, particularly with the introduction of pre-trained models that were fine-tuned for specific tasks following intensive training on large volumes of data. In this paper, the Text-to-Text Transfer Transformer (T5) model which has yielded the best results in a range of NLP tasks is employed to generate an automatic Arabic chatbot. A dataset of 12k Arabic words to improve Arabic Text Transfer Transformer (AraT5), the recently released Arabic language version is used. The effectiveness of the proposed model is evaluated utilizing Bilingual Evaluation Understudy (BLEU) metric. The AraT5 model achieved an impressive BLEU score of 0.72, indicating high-quality language generation and contextual understanding in conversational tasks. This performance not only highlights AraT5’s potential in Arabic chatbot development but also demonstrates its superiority over other state-of-the- art methods that rely on sequence-to-sequence models.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Arabic dialogue generation using AraT5 transformer

  • 期刊论文

  • 发表日期：

    2025年2月3日

查看全文
  • Farida Youness,
  • Ayman Elshenawy,
  • Mohamed Ashraf Madkour

摘要

The use of chatbots in the fields of business and education has greatly expanded in recent years due to their capacity to automate and facilitate communication with users. The creation of chatbots is a difficult but essential task in natural language processing (NLP). Deep learning (DL) approaches have advanced significantly in recent years, particularly with the introduction of pre-trained models that were fine-tuned for specific tasks following intensive training on large volumes of data. In this paper, the Text-to-Text Transfer Transformer (T5) model which has yielded the best results in a range of NLP tasks is employed to generate an automatic Arabic chatbot. A dataset of 12k Arabic words to improve Arabic Text Transfer Transformer (AraT5), the recently released Arabic language version is used. The effectiveness of the proposed model is evaluated utilizing Bilingual Evaluation Understudy (BLEU) metric. The AraT5 model achieved an impressive BLEU score of 0.72, indicating high-quality language generation and contextual understanding in conversational tasks. This performance not only highlights AraT5’s potential in Arabic chatbot development but also demonstrates its superiority over other state-of-the- art methods that rely on sequence-to-sequence models.