Analysis of reviews made to hotel companies using text mining and deep learning methods

With the increasing use of the internet today, the utilization of social networks and the importance of comments on products and services in commercial activities have gained significant attention. Sentiment analysis of comments and texts in virtual environments can be applied in various fields for beneficial purposes. This study aims to enable customers to evaluate hotel reviews as positive, neutral, or negative without having to read them. In this context, customer reviews from Etstur, a hotel reservation platform, were analyzed using deep learning methods. After applying word embedding techniques such as Random, Word2Vec, FastText, GloVe, and BERT to the dataset, deep learning techniques, including the CNN-LSTM hybrid model, GRU model, and DENSE layer, were employed for analysis. The highest accuracy was achieved using the CNN-LSTM deep learning model following the application of the GloVe word embedding technique.