How to apply Machine learning to fight against fake news - A case study in Vietnam
(2022) DABN01 20221Department of Statistics
Department of Economics
- Abstract
- In this study, we aim to apply and compare modified models for fake news detection, namely Two-level Convolutional Neural Network (TCNN) and Two-level Convolutional Neural Network with User Response (TCNN-UR) used in Qian et al. (2018) on social media news in Vietnamese. We also compare our best-fitted model performance to the best model in Hieu et al. (2020) to find how important the corpus size is in solving a text classification problem. We find that the TCNN-UR is more likely to be more robust than its counter- part, i.e., TCNN in terms of model test accuracy. In addition, our best model underperforms when we compare it with an ensemble model with PhoBERT embeddings in Hieu et al. (2020). Although this study has a few weaknesses like... (More)
- In this study, we aim to apply and compare modified models for fake news detection, namely Two-level Convolutional Neural Network (TCNN) and Two-level Convolutional Neural Network with User Response (TCNN-UR) used in Qian et al. (2018) on social media news in Vietnamese. We also compare our best-fitted model performance to the best model in Hieu et al. (2020) to find how important the corpus size is in solving a text classification problem. We find that the TCNN-UR is more likely to be more robust than its counter- part, i.e., TCNN in terms of model test accuracy. In addition, our best model underperforms when we compare it with an ensemble model with PhoBERT embeddings in Hieu et al. (2020). Although this study has a few weaknesses like modest corpus size and inconsistent results while comparing models, our work contributes to confirmation that TCNN models with or without user response features are robust in detecting fake news in a low-resource language like Vietnamese. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9085380
- author
- Pauling, Sofie LU
- supervisor
- organization
- course
- DABN01 20221
- year
- 2022
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- Two-level Convolution Neural Network, User Response, Fake News detection.
- language
- English
- id
- 9085380
- date added to LUP
- 2022-06-08 12:51:12
- date last changed
- 2022-10-10 16:04:47
@misc{9085380, abstract = {{In this study, we aim to apply and compare modified models for fake news detection, namely Two-level Convolutional Neural Network (TCNN) and Two-level Convolutional Neural Network with User Response (TCNN-UR) used in Qian et al. (2018) on social media news in Vietnamese. We also compare our best-fitted model performance to the best model in Hieu et al. (2020) to find how important the corpus size is in solving a text classification problem. We find that the TCNN-UR is more likely to be more robust than its counter- part, i.e., TCNN in terms of model test accuracy. In addition, our best model underperforms when we compare it with an ensemble model with PhoBERT embeddings in Hieu et al. (2020). Although this study has a few weaknesses like modest corpus size and inconsistent results while comparing models, our work contributes to confirmation that TCNN models with or without user response features are robust in detecting fake news in a low-resource language like Vietnamese.}}, author = {{Pauling, Sofie}}, language = {{eng}}, note = {{Student Paper}}, title = {{How to apply Machine learning to fight against fake news - A case study in Vietnam}}, year = {{2022}}, }