Text Mining Data - SET
收藏资源简介：
Emotional classification (valence) in textual data has proved to be central to human experience analysis and natural language processing (NLP). This study implements a text mining model and algorithm - TM-EV (Text Mining for Emotional Valence Analysis) - that determines the impact of emotional valence (EV) shown by undergraduate students in their feedback (n=665860) during the program (pre- and post-course to determine its relationship with the learning outcome and performance. The method is grounded on appraisal theories and component process models (CPM) that study degree of pleasantness or goal achievement as an effect of valence judgements. The model (TM-EV) identifies top terms in the students’ data using Corpus feature selection and Term document matrix libraries in R software. It further utilizes the EV scores (quantified data) extracted from the (textual) data to statistically test the association and effect it has with the Evaluation periods and Academic level of the students. Data analysis was done using Sentiment Analysis libraries (sentimentr, syuzhet, pander) in R, and Statistical Analysis methods (Multiple Linear Regression, ANOVA, ANCOVA) in IBM SPSS v30. The results show that individually the Evaluation periods and Academic level do not directly impact EV scores of the students (p>0.05), but a combined interaction effect of the two factors impacts the EV scores (p=0.003). The paper sheds light on the pedagogical and socio-technical implications of the study’s findings toward achieving improved learning outcomes and sustainable educational practices.
文本数据中的情感效价分类（valence）已成为人类体验分析与自然语言处理（Natural Language Processing，NLP）领域的核心研究方向。本研究构建了一款文本挖掘模型与算法——TM-EV（情感效价分析文本挖掘模型，Text Mining for Emotional Valence Analysis），旨在分析本科生在项目周期内（课程前与课程后）的反馈数据（样本量n=665860）中所体现的情感效价（Emotional Valence，EV），以明确其与学习成果及学业表现的关联。该方法基于评价理论与成分过程模型（Component Process Model，CPM）构建，后者将愉悦度与目标达成程度作为效价判断的结果展开研究。TM-EV模型借助R语言软件中的语料库特征选择与词项-文档矩阵相关库，从学生反馈数据中提取核心词项。该模型进一步利用从文本数据中提取的情感效价量化得分，对学生的评估阶段、学业水平与情感效价之间的关联及影响进行统计学检验。数据分析工作依托R语言的情感分析库（sentimentr、syuzhet、pander）以及IBM SPSS v30中的统计学分析方法（多元线性回归、方差分析ANOVA、协方差分析ANCOVA）完成。研究结果显示，评估阶段与学业水平二者单独均不会对学生的情感效价得分产生直接影响（p>0.05），但二者的联合交互效应可显著影响学生的情感效价得分（p=0.003）。本研究结果可为优化学习成果及践行可持续教育实践提供教学论与社会技术层面的启示。
