Ensemble of Artificial Neural Networks for Aspect Based Sentiment Analysis

Aspect Based Sentiment Analysis is a natural language processing task. The goal of this task is to extract sentiments expressed in online reviews about different aspects of a certain product or service, in order to be further analyzed and aggregated. This paper intends to present the system we developed for solving this task. We used a method consisting of an ensemble of classifiers built using deep learning strategies. It also makes use of the performance and advantages of pretrained word embeddings from the ConceptNet semantic network. We tested two different networks architectures: a recurrent network and a convolutional network. The paper also analyses other top systems architectures from the international workshop on Semantic Evaluation (SemEval-2016) Task 5: Aspect Based Sentiment Analysis. We compare their results and methods with the results provided by our system, using a restaurant reviews dataset provided by the workshop. The results obtained by our method exceed the ones obtained by the presented systems.