Deep Learning and Word Embeddings for Tweet Classification for Crisis Response
This work addresses tweet classification for crisis responders, but it is incremental as it shows comparable results with existing methods.
The paper investigated whether general-purpose word embeddings like GloVe could replace domain-specific ones for tweet classification in crisis response, finding that Bi-LSTM with GloVe achieved the best performance at 62.04% F1 score.
Tradition tweet classification models for crisis response focus on convolutional layers and domain-specific word embeddings. In this paper, we study the application of different neural networks with general-purpose and domain-specific word embeddings to investigate their ability to improve the performance of tweet classification models. We evaluate four tweet classification models on CrisisNLP dataset and obtain comparable results which indicates that general-purpose word embedding such as GloVe can be used instead of domain-specific word embedding especially with Bi-LSTM where results reported the highest performance of 62.04% F1 score.