Sentiment Analysis (SA) employs Natural Language Processing (NLP) techniques in order to infer emotional states and subjective information contained in texts. Generally, previously trained machine learning models are used to identify the polarity of an opinion concerning a given target (e.g. film, book, etc.). Therefore, engineering features in order to create the training set for the learning model is a central task in SA problems. Additionally, finding properly labeled datasets for NLP models containing non-English text is a big challenge. Thus, we aim to contribute to SA in texts written in Brazilian Portuguese (PtBR) by validating the use of ConvNet, a convolutional neural network (CNN) that works with character-level inputs, in analyzing the polarity of product reviews in PtBR. The results obtained from our experiments confirm the model’s efficiency.
CITATION STYLE
da Silva, R. P., Santos, F. A. O., do Nascimento, F. B., & Macedo, H. T. (2018). Cross-Language Approach for Sentiment Classification in Brazilian Portuguese with ConvNets. In Advances in Intelligent Systems and Computing (Vol. 738, pp. 311–316). Springer Verlag. https://doi.org/10.1007/978-3-319-77028-4_42
Mendeley helps you to discover research relevant for your work.