Loading...

Formality Style Transfer Using Deep Neural Network

Ebrahimi, Fatemeh | 2022

113 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55432 (31)
  4. University: Sharif University of Technology
  5. Department: Languages and Linguistics Center
  6. Advisor(s): Sameti, Hossein
  7. Abstract:
  8. Formality style transfer, in other words, automatic transfering style of informal text to formal and vice versa, means changing the style and form of a sentence without changing its content. With the increasing progress of deep neural networks, the formality style transfer in other languages has attracted the attention of other researchers and has made significant progress in natural language processing tasks. Due to the availability of parallel data in the English language, the task of style transfer has been approached and designed basically in the framework of the "encoder-decoder" architecture of neural networks. However, due to the lack of parallel datasets in the Persian language, this task has created a gap in natural language processing applications and has not been considered. The goal of this research is to deploy an automatic system for transferring the style of informal to formal texts in the Persian language using deep neural networks. This system is expected to provide its formal equivalent for each informal sentence. To design such a system primarily, we need a parallel corpus of informal-formal sentences, which was developed by designing a crowdsourcing platform to collect informal to formal parallel data. A semi-supervised approach with consistency training technique has been utilized for this formality style transfer model to obtain the best results. This approach has been able to achieve significant performance compared to the baseline models by using the consistency regularizer and utilizing data perturbation techniques. According to our evaluations and experiments which were carried out in this research, the proposed approach has obtained an 85.3 for Bleu score, 90.66 classification accuracy, and 3.83 for human evaluation scores, which shows about 18.3 to 21% improvement compared to Other baseline methods.

  9. Keywords:
  10. Natural Language Processing ; Persian Language ; Deep Neural Networks ; Consistency Regularizer ; Automatic Text Transfer ; Formality Style Transfer ; Colloquial Texts

 Digital Object List

 Bookmark

No TOC