Topic modeling of journals using Latent Dirichlet Allocation

Creative Commons License

Çakar M., Kartal Y., Gülbandılar E.

4th ICITTBT International Conference , Tirane, Arnavutluk, 30 - 31 Mayıs 2024, ss.42

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Tirane
  • Basıldığı Ülke: Arnavutluk
  • Sayfa Sayıları: ss.42
  • Eskişehir Osmangazi Üniversitesi Adresli: Evet

Özet

As a result of the scientific studies, the number of research journals has increased, resulting in a deep research pool that cannot be thoroughly examined. So, it becomes difficult to reach the desired information, and the time spent is considerably increasing. To solve these problems, it is essential to create Natural Language Processing techniques to facilitate the examination and provide more accessible results. This study aims to examine the research in the journals and determine the subjects the journal covers. For this goal, the texts of the gathered articles were cleaned. Bigrams and trigrams were obtained. Finally, a topic identification study was carried out using the LDA model for 30 topics. As a result of the study, different parameters are effective in LDA applications. It is seen that selecting appropriate LDA parameters may achieve more meaningful results, various pre-processes need to be applied, and a text in a different language affects the performance.