Abstract

Clustering Tiny Tales

Ankita Nandy, Anahit Tahmasyan Bal

Volume: 13 Issue: 4 2023

Abstract:

Social media gives its users an open space to express themselves, which becomes the fodder for numerous research works aimed at understanding the opinion, behavior, and attitudes of the users. While textual analysis was a domain reserved for human eyes, advancement in machine learning has made the analyses of the plethora of tweets, posts, reviews et cetera automated. Such texts often include an explicit mention or reference to the object of interest. On the other hand, creative pieces such as stories, often convey the thought in an implicit manner, and have witnessed limited experimentation with machine learning driven techniques. This work assembles a corpus of Terribly Tiny Tales, a social media presence publishing short stories, originally limited to 140 characters. A manual thematic analysis is followed by an attempt to obtain such relevant clusters through short text clustering techniques of Top2Vec and BERTopic, where the latter is found to generate more meaningful clusters. The quality and validity of a document cluster are based on human understandability; thus, some subjectivity is inherent, and there is plentiful room for further research. Cheap and quality, replica watches for sale online with best movements are suited for men and women.

Buy rolex copy watches sales with high quality on sale.

Buy hot-selling top quality cartier replica watches from online store. Fast shipping. Money back guaranty.

DOI: http://doi.org/10.37648/ijrssh.v13i04.005

Back Download