Abstract:
The assembly of poems is increasing day by day on the internet. A prodigious amount of data
sets are available on the Internet. However, labeling poems is a very important task. The work
in this paper is aimed to find a tagging solution using Bidirectional Long Short-Term
Memory Recurrent Neural Network (BLSTM-RNN) appeared to be very effective for
modeling sequential data. To improve the specific functions cautiously optimal for each
task, our solution only uses a single set of task-independent features. Utilizing task-specific
information and advanced feature engineering, our proposal delivers almost state-of-the-art performance in predicting tagging tasks.
