Friedrich Nietzsche at SemEval-2023 Task 4: Detection of Human Values from Text Using Machine Learning

Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 2179–2183 July 13-14, 2023 ©2023 Association for Computational Linguistics Friedrich Nietzsche at SemEval-2023 Task 4: Detection of Human Values from Text Using Machine Learning Abdul Jawad Mohammed M.S. Computer Science Purdue University Fort Wayne, Indiana mohaa05@pfw.edu Sruthi Sundharram M.S. Computer Science Purdue University Fort Wayne, Indiana sunds01@pfw.edu Sanidhya Sharma M.S. Computer Science Purdue University Fort Wayne, Indiana shars05@pfw.edu Abstract Literature permeates almost every facet of our lives, whether through books, magazines, or in- ternet articles. Moreover, every piece of written work contains ideas and opinions that we tend to relate to, accept or disregard, the debate over, or enlighten ourselves. However, the existence of subtle themes that are difficult to discern inspired us to utilize four machine learning al- gorithms: Decision Trees, Random Forest, Lo- gistic Regression, and Support Vector Classifier (SVC) to aid in their detection. Trained on the ValueEval data set as a multi-label classifica- tion problem, the supervised machine learning models did not perform as well as expected, with F1 metrics hovering from 0.0 to 0.04 for each value. Noting this, our paper discusses our approach’s limitations and weaknesses. 1 Introduction The creation of language remains one of human- ity’s greatest achievements, through which con- cepts, knowledge, stories, memories and emotions are encapsulated into words and conveyed across generations. Among the multitude of ideas con- tained within the ever-growing expanse of written texts, whether it be news articles, novels, maga- zines or philosophical and religious scripture, there exists a commonality: The commentary on hu- manistic archetypes of freedom, creativity, self- preservation, etc. upon which rational beings base their actions and thoughts on. These notions are expressed textually in varying degrees, sometimes apparent while at other times too subtle and dif- ficult for the layman to recognize. For example, advertisements and political campaigns try to iden- tify with potential customers and voters by openly appealing to their human values such as religious beliefs and concerns about nature and the environ- ment. At other times, however, ambiguous and obscure literature could foster misunderstandings, misinterpretations, and even conflict. Therefore, detecting and determining human values in texts could serve a potent role in many Natural Language Processing applications, including the categoriza- tion of text information for tag-based recommen- dation systems, opinion/stance detection in arti- cles, and analysis of complex literature written by poets and philosophers. As part of the SemEval 2023 ValueEval task (Kiesel et al., 2023), we seek to explore the effectiveness of supervised earning classifiers for semantic analysis through the utiliza- tion of algorithms like Decision Trees, Logistic Regression, Support Vector Machines, and Ran- dom Forest in the extraction of implicit themes within the literature. With most research on seman- tic analysis being conducted using more powerful transformer/neural network based models, our pa- per chose to focus on more traditional machine learning techniques for their potential in being inte- grated with lightweight, portable recommendation systems and software based on thematic analysis of literature. Due to its vast …

Content truncated to 500 words