A Complete Text-Processing Pipeline for Business Performance Tracking

Natural text processing is amongst the most researched domains because of its varied applications. However, most existing works focus on improving the performance of machine learning models instead of applying those models in practical business cases. We present a text processing pipeline that enables business users to identify business performance factors through sentiment analysis and opinion summarization of customer feedback. The pipeline performs fine-grained sentiment classification of customer comments, and the results are used for sentiment trend tracking process. The pipeline also performs topic modelling in which key aspects of customer comments are clustered using their co-relation scores. The results are used to produce abstractive opinion summarization. The proposed text processing pipeline is evaluated using two business cases in the food and retail domains. The performance of the sentiment analysis component is measured using mean absolute error (MAE) rate, root mean squared error (RMSE) rate, and coefficient of determination.