Get our free extension to see links to code for papers anywhere online!

Chrome logo  Add to Chrome

Firefox logo Add to Firefox


BERT-Based Combination of Convolutional and Recurrent Neural Network for Indonesian Sentiment Analysis

Add code

Nov 10, 2022
Hendri Murfi, Syamsyuriani, Theresia Gowandi, Gianinna Ardaneswari, Siti Nurrohmah


Share this with someone who'll enjoy it:


Sentiment analysis is the computational study of opinions and emotions ex-pressed in text. Deep learning is a model that is currently producing state-of-the-art in various application domains, including sentiment analysis. Many researchers are using a hybrid approach that combines different deep learning models and has been shown to improve model performance. In sentiment analysis, input in text data is first converted into a numerical representation. The standard method used to obtain a text representation is the fine-tuned embedding method. However, this method does not pay attention to each word's context in the sentence. Therefore, the Bidirectional Encoder Representation from Transformer (BERT) model is used to obtain text representations based on the context and position of words in sentences. This research extends the previous hybrid deep learning using BERT representation for Indonesian sentiment analysis. Our simulation shows that the BERT representation improves the accuracies of all hybrid architectures. The BERT-based LSTM-CNN also reaches slightly better accuracies than other BERT-based hybrid architectures.

* 15 pages 


   Access Paper Source



Share this with someone who'll enjoy it: