Optimizing sentiment strength detection for classification for airline opinion texts

Abstract

Short text is rapidly expanding, and it is critical to understand why. Unlike traditional documents, short text is short and sparse, making machine learning and text mining algorithms difficult to apply. The process of identifying and categorizing ideas expressed in a text into positive, negative, and neutral polarities is referred to as sentiment analysis. The selection of features for machine learning is critical. The evaluation of search engines is a critical topic in the field of information retrieval. Researchers can determine whether their new algorithms add value or should be scrapped by effectively evaluating their algorithms. Organizations create cutting-edge methods for assessing their performance. Social tags have become an important source of information in web 2.0 for making personalized recommendations by identifying user interests and preferences. The purpose of this research is to systematically represent the actual topics of tags, the content of items, and user interest in specific topics via popular tags based on the content of the tags, items, and topics. The goal of this paper is to propose an innovative method for identifying user topic interests using popular tags. This paper proposes a method for extracting the main features of short text at various granularity levels. One aspect of how the content of items is represented in this paper is through the use of popular tags, but the representation of user interests is also discussed. By analysing a small dataset of short text classifications, we were able to compare our proposed method to a state-of-the-art baseline. We reduced the classification error by 20% using two different analysis methods, while the other method reduced it by 16.68%. The original tags of users are used to generate topic interests, from which we identify popular tags. The popular tags are then used to generate topic interests using an autoencoder approach.