Development of a web application for intelligent analysis of customer reviews using a modified seq2seq model with an attention mechanism

Machine learning, and neural networks in particular, are having a huge impact on business and marketing by providing convenient tools for analytics and customer feedback. The article proposes an intelligent analysis of customer feedback based on the use of a modified seq2seq deep learning model. Since the basic seq2seq model has a significant disadvantage – the inability to concentrate on the main parts of the input sequence, the results of machine learning may give an inadequate assessment of customer feedback. This disadvantage is eliminated by means of a model proposed in the work called the “attention mechanism”. The model formed the basis for the development of a web application that solves the problem of flexible interaction with customers by parsing new reviews, analyzing them and generating a response to a review using a neural network.