Emoji-Integrated Polyseme Probabilistic Analysis Model: Sentiment Analysis of Short Review Texts on Library Service Quality

Abstract

It is a great challenge to understand user evaluation of library service quality based on short review texts. This is because short texts are limited in length and lack context support. What is worse, the polysemes and emojis in short texts make the literal emotions of these texts rather ambiguous and variable. The variability is often overlooked in previous research on service quality evaluation, which reduces the accuracy of automatic analysis methods. Considering the effects of polysemes and emojis in short texts, this paper introduces probabilistic linguistic term sets (PLTS) and support vector machine (SVM) to establish a framework for emotional classification of library service quality (FECLSQ). Every word and emoji were converted into the corresponding PLTS to depict the probability of the word/emoji belonging to each sentiment polarity, making short text sentiment analysis more accurate. Through supervised learning of corpuses, the authors established the PLTSs of polysemes, and context sentiment weight dictionary (CSWD), and coupled them with the FECLSQ for sentiment analysis and application of text sets with various themes. The proposed approach was utilized to correctly evaluate library service quality.