Shrivas, Rohit Kumar (2021) To what extent NLP with RNN and Transformer Based Deep Neural Network can be used to classify Insincere questions on Quora. Masters thesis, Dublin, National College of Ireland.
|
PDF (Master of Science)
Download (1MB) | Preview |
|
|
PDF (Configuration manual)
Download (1MB) | Preview |
Abstract
This research project adopts a novel approach of utilizing BERT based models for question classification application of NLP. The performance of these models are then evaluated using F1 score and compared with the performance of RNN based Deep Learning models with and without Attention layer. The DistilBERT model outperformed every model implemented in the research and achieved an F1 score of 94.87% which is 7.06% improvement over previous studies. Among the RNN based Deep Learning models, Model-6 (Bi-directional GRU + Bi-directional LSTM + Attention layer) performed the best and achieved an F1 score of 63.53%.
Actions (login required)
![]() |
View Item |