Aspect Level Opinion Mining for Hotel Text Reviews using Traditional Machine Learning Approaches

Abstract

The customers intend to write their opinion about the hotel they visit in order to provide suggestion, complaint, or appreciate the services. These opinions have been expressed in terms of text on the online platforms. The hotel industries need to process and analyze these opin- ions to improve their  business  in  the  right  direction.  The challenges such as computational efficiency to process a huge size data ore of opinions and ability  to  learn  from  the  continuously  generated  data. This article proposes a machine learning based opinion mining method that captures the intention of the  customer  as  well  as  the  opinions with respect to the  specific  aspects  of  the  hotel  entities.  The  pro- posed method  has  a  three  stage  pipeline  that  includes  pre-processing of the textual opinions, feature extraction from the pre-processed opin- ions,  and  classification  of  the  features  values  into  its  correct  rating and various sub-ratings.  The experiments have been carried  out  on 4.5M reviews of HotelRec dataset. The results obtained through the proposed method has outperformed the existing baseline methods.