NL-FIIT at IEST-2018: Emotion Recognition utilizing Neural Networks and Multi-level Preprocessing
| Autor: | Mária Bieliková, Marián Šimko, Peter Lacko, Michal Farkas, Samuel Pecar |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Artificial neural network
Computer science Speech recognition 010401 analytical chemistry 010501 environmental sciences 01 natural sciences 0104 chemical sciences Task (project management) Preprocessor F1 score Word (computer architecture) Dropout (neural networks) Sentence 0105 earth and related environmental sciences |
| Zdroj: | WASSA@EMNLP |
| DOI: | 10.18653/v1/w18-6231 |
| Popis: | In this paper, we present neural models submitted to Shared Task on Implicit Emotion Recognition, organized as part of WASSA 2018. We propose a Bi-LSTM architecture with regularization through dropout and Gaussian noise. Our models use three different embedding layers: GloVe word embeddings trained on Twitter dataset, ELMo embeddings and also sentence embeddings. We see preprocessing as one of the most important parts of the task. We focused on handling emojis, emoticons, hashtags, and also various shortened word forms. In some cases, we proposed to remove some parts of the text, as they do not affect emotion of the original sentence. We also experimented with other modifications like category weights for learning and stacking multiple layers. Our model achieved a macro average F1 score of 65.55%, significantly outperforming the baseline model produced by a simple logistic regression. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|