PRFashion24: A Dataset for Sentiment Analysis of Fashion Products Reviews in Persian

Overview

• The PRFashion24 dataset is a comprehensive Persian dataset collected from various online fashion stores, spanning from April 2020 to March 2024.
• The dataset contains 767,272 reviews, making it the first of its kind in the Persian language that encompasses diverse categories within the fashion industry.
• The study aimed to harness deep learning techniques, specifically Long Short-Term Memory (LSTM) networks and a combination of Bidirectional LSTM and Convolutional Neural Network (BiLSTM-CNN), to analyze and reveal sentiments towards online fashion shopping.

Plain English Explanation

The PRFashion24 dataset is a collection of over 767,000 reviews from Persian-language online fashion stores, gathered over the course of four years. This is the first dataset of its kind to provide such a comprehensive look at how people feel about shopping for fashion online in the Persian language.

The researchers used two different deep learning models, LSTM and BiLSTM-CNN, to analyze the sentiment expressed in these reviews. LSTM is a type of artificial neural network that is well-suited for processing sequential data, like text. BiLSTM-CNN combines a Bidirectional LSTM, which can analyze text from both directions, with a Convolutional Neural Network, which is good at identifying patterns in data.

The LSTM model was able to accurately predict the sentiment (positive or negative) of the reviews 81.23% of the time. The BiLSTM-CNN model performed even better, reaching an accuracy of 82.89%. These findings provide valuable insights into the attitudes and opinions of Persian-speaking online fashion shoppers, which tend to be more positive overall.

By making this dataset and the optimized models publicly available on GitHub, the researchers hope to contribute to the understanding of sentiment analysis in the Persian-language context, as well as inspire further research in this area.

Technical Explanation

The researchers collected the PRFashion24 dataset from various online Persian-language fashion stores, spanning a period of four years from April 2020 to March 2024. The dataset consists of 767,272 reviews, making it the first of its kind to provide such a comprehensive coverage of the Persian fashion industry.

To analyze the sentiment expressed in these reviews, the researchers employed two deep learning techniques: Long Short-Term Memory (LSTM) networks and a combination of Bidirectional LSTM and Convolutional Neural Network (BiLSTM-CNN). LSTM is a type of recurrent neural network that is well-suited for processing sequential data, such as text, and can capture long-term dependencies. The BiLSTM-CNN model combines the strengths of a Bidirectional LSTM, which can analyze text from both directions, and a Convolutional Neural Network, which is adept at identifying patterns in data.

The LSTM model achieved an accuracy of 81.23% in predicting the sentiment (positive or negative) of the reviews. The BiLSTM-CNN model performed even better, reaching an accuracy of 82.89%. These results demonstrate the effectiveness of deep learning techniques in analyzing sentiment in Persian-language fashion reviews.

By making both the optimized models and the PRFashion24 dataset publicly available on GitHub, the researchers aim to contribute to the understanding of sentiment analysis in the Persian-language context and inspire further research in this area.

Critical Analysis

The PRFashion24 dataset and the associated deep learning models provide valuable insights into the sentiments of Persian-speaking online fashion shoppers. However, there are a few areas that could be explored further.

One potential limitation is the representativeness of the dataset. While it is the first of its kind in the Persian language, it is still limited to a specific time frame and may not capture the full range of opinions and experiences across all demographics and geographic regions. Expanding the dataset to include a more diverse set of reviews over a longer period could help strengthen the findings.

Additionally, the study focuses solely on sentiment analysis, but it might be interesting to explore other aspects of the data, such as the relationships between product categories, brand preferences, and customer behavior. Incorporating these elements into the analysis could provide a more comprehensive understanding of the Persian fashion industry.

Furthermore, the researchers could consider applying other deep learning architectures or hybrid approaches, such as the use of attention mechanisms or multimodal models, to see if they can further improve the sentiment analysis performance.

Overall, the PRFashion24 dataset and the deep learning models presented in this study represent a valuable contribution to the field of sentiment analysis in the Persian-language context. By making the dataset and models publicly available, the researchers have opened the door for further research and exploration in this area.

Conclusion

The PRFashion24 dataset and the associated deep learning models provide a comprehensive analysis of sentiment in Persian-language online fashion reviews. With over 767,000 reviews spanning diverse fashion categories, this dataset is a significant step forward in understanding the attitudes and opinions of Persian-speaking online shoppers.

The LSTM and BiLSTM-CNN models developed in this study demonstrate the effectiveness of deep learning techniques in sentiment analysis, achieving accuracies of 81.23% and 82.89%, respectively. These findings not only contribute to the understanding of sentiment analysis in the Persian-language context but also have the potential to inform businesses and policymakers about the preferences and concerns of Persian-speaking fashion consumers.

By making the dataset and optimized models publicly available on GitHub, the researchers have opened the door for further exploration and innovation in this field. This research sets the stage for future studies that could delve deeper into the relationships between product categories, brand preferences, and customer behavior, as well as the application of more advanced deep learning architectures to enhance sentiment analysis in the Persian-language fashion domain.