School of Biological Sciences, Nanyang Technological University, Singapore, Singapore.
College of Engineering, Nanyang Technological University, Singapore, Singapore.
PLoS One. 2023 Jan 12;18(1):e0274299. doi: 10.1371/journal.pone.0274299. eCollection 2023.
Sentiment Analysis (SA) is a category of data mining techniques that extract latent representations of affective states within textual corpuses. This has wide ranging applications from online reviews to capturing mental states. In this paper, we present a novel SA feature set; Emotional Variance Analysis (EVA), which captures patterns of emotional instability. Applying EVA on student journals garnered from an Experiential Learning (EL) course, we find that EVA is useful for profiling variations in sentiment polarity and intensity, which in turn can predict academic performance. As a feature set, EVA is compatible with a wide variety of Artificial Intelligence (AI) and Machine Learning (ML) applications. Although evaluated on education data, we foresee EVA to be useful in mental health profiling and consumer behaviour applications. EVA is available at https://qr.page/g/5jQ8DQmWQT4. Our results show that EVA was able to achieve an overall accuracy of 88.7% and outperform NLP (76.0%) and SentimentR (58.0%) features by 15.8% and 51.7% respectively when predicting student experiential learning grade scores through a Multi-Layer Perceptron (MLP) ML model.
情感分析（SA）是一种数据挖掘技术，用于从文本语料库中提取情感状态的潜在表示。它的应用范围广泛，从在线评论到捕捉心理状态。在本文中，我们提出了一种新的情感分析特征集；情感方差分析（EVA），它可以捕捉到情感不稳定的模式。我们将 EVA 应用于从体验式学习（EL）课程中获取的学生日志，发现 EVA 可用于分析情感极性和强度的变化，从而预测学习成绩。作为一个特征集，EVA 与各种人工智能（AI）和机器学习（ML）应用程序兼容。虽然是在教育数据上进行评估，但我们预计 EVA 将在心理健康分析和消费者行为应用中有用。EVA 可在 https://qr.page/g/5jQ8DQmWQT4 上获取。我们的研究结果表明，EVA 通过多层感知器（MLP）ML 模型预测学生体验式学习成绩时，其整体准确率为 88.7%，比自然语言处理（NLP）（76.0%）和 SentimentR（58.0%）特征分别高出 15.8%和 51.7%。
