EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Comparative analysis of unidirectional and bidirectional LSTM networks for sentiment classification in film reviews using TensorFlow

Diego Andrés Castellano Carvajal, Oriana Alexandra López - Bustamante, Jorge Enrique Herrera - Rubio

Mundo FESC · 01.01.2026 · 16 (34) · 100-110

DOI: 10.61799/2216-0388.2100 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Mundo FESC, online erschienen am 01.01.2026. Verfasst von Diego Andrés Castellano Carvajal und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Due to the rapid increase in the quantity and size of textual data generated daily on various digital platforms, sentiment analysis using deep learning methods has gained popularity in natural language processing applications. Recurrent neural networks (RNNs) have been successfully used to model the semantic and temporal dependencies present in a text sequence thanks to their ability to learn sequential information. In this article, the authors compare the performance of unidirectional and bidirectional long-term memory (LSTM) networks for binary sentiment classification of IMDb movie reviews. The authors used various methods to preprocess the text data and convert it into a sequence of vectors using TensorFlow, as well as trainable embeddings and regularization using Early Stopping. The study concludes that the bidirectional LSTM model achieved an accuracy of 99.87%, while the unidirectional LSTM model achieved an accuracy of 97.01%, demonstrating that the bidirectional LSTM model exhibits greater convergence stability, both during and after training, than the unidirectional version. Furthermore, the authors provide additional evidence, in the form of graphs illustrating the differences between the two models, demonstrating that bidirectional contextual processing significantly contributes to improving semantic language representations, as well as overall performance on sentiment analysis tasks.

Zitieren

APA

Carvajal, D. A. C., Bustamante, O. A. L., & Rubio, J. E. H. (2026). Comparative analysis of unidirectional and bidirectional LSTM networks for sentiment classification in film reviews using TensorFlow. Mundo FESC, 16(34), 100-110. https://doi.org/10.61799/2216-0388.2100

Vancouver

Carvajal DAC, Bustamante OAL, Rubio JEH. Comparative analysis of unidirectional and bidirectional LSTM networks for sentiment classification in film reviews using TensorFlow. Mundo FESC. 2026;16(34):100-110. doi:10.61799/2216-0388.2100

BibTeX
@article{Carvajal2026,
  author = {Carvajal, Diego Andrés Castellano and Bustamante, Oriana Alexandra López - and Rubio, Jorge Enrique Herrera -},
  title = {Comparative analysis of unidirectional and bidirectional LSTM networks for sentiment classification in film reviews using TensorFlow},
  journal = {Mundo FESC},
  year = {2026},
  volume = {16},
  number = {34},
  pages = {100--110},
  doi = {10.61799/2216-0388.2100}
}
Kostenlos starten →