Determining the ability of artificial intelligence to establish authorship of artistic ukrainian texts using significant fragments

Authors

  • Ivanov Oleksandr
  • Skalozub Vladislav
  • Horiachkin Vadym
  • Shynkarenko Viktor

DOI:

https://doi.org/10.34185/1562-9945-5-148-2023-08

Keywords:

definition of authorship, natural language text, artificial intelligence, generative language models, ChatGPT, Bing bot, Skype.

Abstract

Artificial intelligence is becoming an integral part of everyday life and profes-sional activity of a person. Bing, as an intelligent search system, can serve as a tool for determining the authorship of artistic text in Ukrainian. Bing helps to uncover in-formation about a text fragment and its author, although the search results may be inaccurate or incomplete. This work aims to explore the effectiveness of determining the authorship of ar-tistic texts using modern artificial intelligence tools based on significant fragments of works. Ten Ukrainian authors with a rich body of artistic works, reflecting various as-pects of Ukrainian culture and history, were selected for the experiment. Random fragments up to 500 words in length were selected from various works of these au-thors. An experiment was conducted to determine the authorship of 360 fragments. Using the Python programming language and the skpy package, software was created that sends queries and receives responses from the Bing bot embedded in Microsoft Skype. The presence of the author’s name and the corresponding title of the work were checked in the response texts. This work introduces, for the first time, a method of verifying the authorship of Ukrainian-language text fragments using the Bing bot equipped with artificial intelli-gence. A comparative analysis was conducted and experiments were carried out to identify the authorship of significant long fragments. It was found that long fragments allow the author of the artistic Ukrainian text to be determined with high accuracy. Ivan Franko has the highest percentage of re-sponses where the author’s name and the title of the work were mentioned - 87%. The proposed hypothesis regarding the effectiveness of artificial intelligence in establishing authorship of works has not been confirmed. Artificial intelligence has slightly lower efficiency than expected, which indirectly exposes its means of opera-tion. Namely, when establishing authorship, a sequential research comparison of the proposed fragment with a bank of works that are widespread in the Internet environ-ment is not performed.

References

Chowdhery A. PaLM: Scaling Language Modeling with Pathways, Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts.// https://doi.org/10.48550/arXiv.2204.02311 – 2022.

An unofficial Python library for interacting with the Skype HTTP API. [Electronic resource] – Access mode: https://pypi.org/project/SkPy/.

Bengio Y. «Neural net language models». Scholarpedia. Dr. Yoshua Bengio 3. doi:10.4249/scholarpedia.3881. – 2008 – P. 3881.

Perez E. True few-shot learning with language models. Advances in Neural Information Processing Systems, Ethan Perez, Douwe Kiela, and Kyunghyun Cho. –2021 – P. 34.

Google Bard: How to Use Google's AI Chatbot - How-To Geek. [Electronic re-source] – Access mode: https://www.howtogeek.com/880668/google-bard-how-to-use-googles-ai-chatbot/.

How Bing AI Search Uses Website Content - Search Engine Journal. [Electronic resource] – Access mode: https://www.searchenginejournal.com/how-bing-ai-search-uses-web-content/480643/.

Microsoft’s next-gen Bing uses a ‘much more powerful’ language model than ChatGPT [Electronic resource] – Access mode:

https://www.engadget.com/microsofts-next-gen-bing-more-powerful-language-model-than-chatgpt-182647588.html.

Reinventing search with a new AI-powered Microsoft Bing and Edge, your copilot for the web. [Electronic resource] – Access mode:

https://blogs.microsoft.com/blog/2023/02/07/reinventing-search-with-a-new-ai-powered-microsoft-bing-and-edge-your-copilot-for-the-web/.

Ivanov O. P. Determining the Authorship of a Ukrainian-Language Literary Text by Means of Artificial Intelligence from Ultra-Short Excerpts O. P. Ivanov, V.I. Shynkarenko, V.V. Skalozub, A.A. Kosolapov. Science and Transport Progress № 2(102) – 2023 – P. 45-53.

Shynkarenko V.I. Methods and software for significant indicators determination of the natural language texts author profile V.I. Shynkarenko, I.M. Demydovych. // Prombles in programming/ 3 – 2023 – С. 22-29.

Shynkarenko V.I. A Dual Approach to Establishing the Authority of Technical Natural Language Texts and Their Components V. I. Shynkarenko, I. M. Demidovich, O. S. Kuropiatnyk // Science and Transport Progress № 2(102) – 2023 – С. 71-85.

Downloads

Published

2024-03-20