Do you like (or dislike) “fruit flies like a banana,” but not “time flies like an arrow”? How should humans and computerized systems accurately distinguish between predefined and undefined categories of words used in sentences to reflect emotions and attitudes? Zhu et al. offer a new framework to help current and future automated systems learn in real time and cope with the difficulties in processing frequently confusing semantic words.

The process of recognizing the proven characteristics of emotion clusters is called aspect category sentiment analysis (ACSA). The identification of terms associated with an emotion in a sentence is aspect term sentiment analysis (ATSA). The authors succinctly critique the existing algorithms in the literature, designed to solve ACSA and ATSA issues. In fact, it is not easy to locate emotion clusters from sentences, due to the complexity of distinguishing between the semantic contexts in the uses of words and terms in sentences. Consequently, the authors present an aspect aware learning (AAL) model for overcoming ACSA problems.

The AAL model consists of: (1) a deep learning algorithm for semantically relating words to sentences for emotional categorization; (2) an algorithm capable of recognizing layers of changing emotional behaviors; and (3) two alternative algorithms for assessing the precise use of words and terms in emotional sensitivity investigations.

Experiments were performed with different subjects at various restaurants and locations. The efficiencies of the presented AAL algorithms compare favorably with the existing ones in the literature. I strongly invite all information retrieval specialists, librarians, and deep learning and artificial intelligence (AI) colleagues to read and offer new directions for this new and clumsy area of AI computer applications.