Conversational Artificial Intelligence

4Unsupervised BERT-Based Granular Sentiment Analysis of Literary Work

N. Shyamala Devi^* and K. Sharmila

Department of Computer Science, Vels Institute of Science, Technology & Advanced Studies, Chennai, India

Abstract

Data mining is an important domain that is essential to many industries irrespective of the information they hold. However, modern times entail the need to extract relevant information from immense volumes of data that may be unstructured or structured. Sentiment analysis is a subfield of the mining process and involves the extraction of the perspectives established from individuals or collective entities. The existing methodologies have highlighted various classifiers and combinational processing methods to analyze the sentiments of different platforms. Nonetheless, the proposed methodology pivots to identify the sentiment on the literary work of Shakespeare. The play Hamlet is a popular piece of study, and the extraction of relative sentiments involved is accomplished through natural language processing methods. The sentiments identified in the dialogues of the play are recognized through fine-grained text analysis and implemented through a sequence of phases to preprocess, extract, and vectorize the features involved. An unsupervised training and classification method through the BERT machine learning is then applied to the literary work to label and further classify the data into sectorized clusters.

Keywords: Natural language processing (NLP), BERT, vectorize, ...