Aspect Term Extraction Using Deep Learning Model with Minimal Feature Engineering

With the explosive growth of social media on the Web, opinion mining has been extensively investigated and consists of the automatic identification and extraction of opinions, emotions, and sentiments from text and multimedia data. One of the tasks involved in opinion mining is Aspect Term Extraction (ATE) which aims at identifying aspects (attributes or characteristics) that have been explicitly evaluated in a sentence or a document. For example, in the sentence “The picture quality of this camera is amazing”, the aspect term is “picture quality”. This work proposes POS-AttWD-BLSTM-CRF, a neural network architecture using a deep learning model, and minimal feature engineering, to solve the problem of ATE in opinionated documents. The proposed architecture consists of a BLSTM-CRF classifier that uses the part-of-speech tag (POS tags) as an additional feature, along with a BLSTM encoder with an attention mechanism to allow the incorporation of another relevant feature: the grammatical relations between words. The experiments show that the proposed architecture achieves promising results with minimal feature engineering comparing to the state-of-the-art solutions.