Prototyping a Reusable Sentiment Analysis Tool for Machine Learning and Visualization

Abstract

Evaluating customer satisfaction is very significant in all organizations to get the perspective of users/customers/stakeholders on products and/or services. Part of the data obtained during the evaluation are observations and comments of respondents and these are very rich in insights as they provide information on the strengths as well as the areas needing improvement. As the volume of textual data increases, the difficulty of analyzing them manually also increases. With these concerns, text analytics tools should be used to save time and effort in analyzing and interpreting the data. The textual data being processed in sentiment analysis problems vary in so many ways. For instance, the context of textual data and the language used vary when data are sourced from different locations and areas or fields. Thus, machine learning was utilized in this study to customize text analysis depending on the context and language used in the dataset. This research aimed to produce a prototype that can be used to explore three vectorization techniques and selected machine learning algorithms. The prototype was evaluated in the context of features for the application of machine learning in sentiment analysis. Results of the prototype development and the feedback and suggestions during the evaluation were presented. In future work, the prototype shall be improved, and the evaluators' feedback will be considered.