|
Incentivized comment detection with sentiment analysis on hotel reviews |
|---|---|
| รหัสดีโอไอ | |
| Creator | Niaz Imtiaz |
| Title | Incentivized comment detection with sentiment analysis on hotel reviews |
| Contributor | Toukir Ahmed, Antara Paul |
| Publisher | Asia-Pacific Journal of Science and Technology |
| Publication Year | 2565 |
| Journal Title | Asia-Pacific Journal of Science and Technology |
| Journal Vol. | 27 |
| Journal No. | 3 |
| Page no. | 11 |
| Keyword | Incentivized reviews, Sentiment analysis, Text mining, Machine learning |
| URL Website | https://www.tci-thaijo.org/index.php/APST |
| Website title | https://so01.tci-thaijo.org/index.php/APST/article/view/240435 |
| ISSN | 2539-6293 |
| Abstract | With the enormous platforms currently available, consumers communicate and interconnect online frequently with web users all around the world to share their experiences. Thus, online platforms have become a major source of reviews and opinions about different entities. People travel frequently around the world for different purposes. Seeking good hotels for accommodation is a prime concern. Reviews on hotels from customers help future customers to make decisions about their accommodation and help hotel owners to think about designing customer facilities. However, many online reviews are biased due to different factors. Many hotel owners often come up with attractions like referral rewards, coupons, bonus points, etc. for the reviewers to motivate them to write biased reviews. We have looked at reviews on 100 hotels in the US and found 952 incentivized reviews among 19175 reviews, which is 4.96% of the total number of reviews. A categorization of incentivized reviews is performed as well. Furthermore, hotels are distinguished based on real and incentivized reviews about them. The results are verified using machine learning algorithms. The Random Forest, K-Nearest Neighbor and Support Vector Machine machine learning algorithms are applied to validate the accuracy of our model, and their prediction results are compared. Random Forest outperforms the others with 94.4% prediction accuracy. |
