The goal of this study is to compare the results of Corpus analysis and Sentiment AI analysis of emotional attitude of taste terms in English. We first show that the meaning of taste terms is multidimensional: In the semantic descriptive dimension, (a) a literal meaning of the taste; (b) a figurative meaning that has been extended to the state of affairs; and, in the semantico-pragmatic evaluative dimension, (c) a speaker’s positive or negative emotional attitude toward the content. In exploring the collocation patterns of basic taste terms with other emotionally charged elements, extracted from Corpus of Contemporary American English (COCA), we suggest there is systematic interaction between various evaluative items. Further, we compare this result of the COCA analysis with the prediction by Sentiment AI and discuss the advantages and limitations of a current Sentiment AI model. Theoretical implications include the following: (i) These big data analyses allow us to uncover subtle connotational differences with respect to positive or negative sentience; and (ii) they support the notion of multidimensionality in meaning.
1. Introduction
2. The Landscape of Taste Terms in American English
3. Corpus-based Sentiment Analysis of Taste Terms
4. Sentiment AI Analysis of Taste Terms
5. Conclusion
(0)
(0)