- Title
- Modeling Fake News Detection Using BERT-CNN-BiLSTM Architecture
- Creator
- Alghamdi, Jawaher; Lin, Yuqing; Luo, Suhuai
- Relation
- Fifth International Conference on Multimedia Information Processing and Retrieval (MIPR 2022). Proceedings of the Fifth International Conference on Multimedia Information Processing and Retrieval (MIPR 2022) ( 2-4 August, 2022) p. 354-357
- Publisher Link
- http://dx.doi.org/10.1109/MIPR54900.2022.00069
- Publisher
- Institute of Electrical and Electronics Engineers (IEEE)
- Resource Type
- conference paper
- Date
- 2022
- Description
- The work presented in this paper explores a deep 6-way multi-class classifier based on the existing pre-trained Bidirectional Encoder Representations from Transformers (BERT) for classifying a statement into six fine- grained categories offake news. Our framework consists of three main components: (1) BERTbase is used to encode and represent the text data into meaningful vectors, which are then fed into a CNN followed by a max-pooling layer for feature map reduction; (2) the metadata of the statement is encoded using another CNN to capture local patterns, and then the output is fed into a BiLSTM to extract more contextual features; and (3) these two components are concatenated and passed to a fully connected layer for classification. Our findings reveal the importance of conducting a feature selection ap-proach. We believe a stronger focus should be placed on pre-processing the data in question by selecting those relevant features (e.g., credit history) since some other features are rather confusing to the classifier, leading to performance degradation. The effectiveness of the proposed framework is evaluated on a benchmark dataset, and the obtained results demonstrate that the proposed framework is very effective, if not superior, compared with the state-ofthe-art fake news detection approaches.
- Subject
- bit error rate; information processing; metadata; feature extraction; transformers; history
- Identifier
- http://hdl.handle.net/1959.13/1492206
- Identifier
- uon:53271
- Identifier
- ISBN:9781665495486
- Language
- eng
- Reviewed
- Hits: 926
- Visitors: 925
- Downloads: 0
| Thumbnail | File | Description | Size | Format |
|---|