Created a tool that can categorizes the Bengali news headlines into six category (National, Politics, International, Sports, Amusement, IT) using deep recurrent neural network.
A dataset of 0.13 Million news headlines is used which has been forked from Eftekhar-Hossain's Git Repository.
News headlines are from different Bengali online news portals such as Dainik Jugantor, Dainik Ittefaq, Dainik Kaler Kontho and so on.
Word embeeding feature represtations technique (Bengali Glove 300d) from sagorbrur Repository is used for extracting the semantic meaning of the words.
A deep learning model has been built by using a 1 layer Unidirectional Recurrent Neural Network LSTM.
Finally, the model performance is evaluated using various evaluation measures such as confusion matrix, accuracy , precision, recall and f1-score.
Developement Environment : Jupyter Notebook
Python Version : 3.8
Framework and Packages : PyTorch, Scikit-Learn, Pandas, Numpy, Matplotlib, Seaborn