Aspect Based Opinion Mining Using Global Vectors And Recurrent Connection

Abstract

Aspect level based opinion mining is the key technique to extract multiple objective information from the text documents written in natural languages. This process helps various businesses to gain actionable sights to further improve their business. The recent abrupt in usage of internet has given the chance to the customer to report good and bad aspects of their consumption. These online posts of a customer on any platform impacts the business as well as market so, the task of aspect level based opinion mining has become of great importance. This article proposes a deep learning based generic method for opinion mining of online customer text reviews  from  the  HotelRec  dataset.  The proposed  method  adopts a GloVe word embedding to extract the features from the natural text. The extracted features are then fed to the recurrently connected gated recurrent unit cell for classifying a review into one  overall  rating  and eight key aspect level sub-ratings. This study has compared various state- of-the-art text embedding techniques to extract the features from the textual reviews. The performance obtained from the proposed method has outperformed the existing state-of-the-art methods in this regard.