Extracting fine-grained information from social media is traditionally a challenging\ntask, since the language used in social media messages is usually informal,\nwith creative genre-specific terminology and expression. How to handle\nsuch a challenge so as to automatically understand the opinions that people\nare communicating has become a hot subject of research. In this paper, we\naim to show that leveraging the pre-learned knowledge can help neural network\nmodels understand the creative language in Tweets. In order to address\nthis idea, we present a transfer learning model based on BERT. We fine-turned\nthe pre-trained BERT model and applied the customized model to two downstream\ntasks described in SemEval-2018: Irony Detection task and Emoji Prediction\ntask of Tweets. Our model could achieve an F-score of 38.52 (ranked\n1/49) in Emoji Prediction task and 67.52 (ranked 2/43) and 51.35 (ranked\n1/31) in Irony Detection subtask A and subtask B. The experimental results\nvalidate the effectiveness of our idea.
Loading....