From Text to Insight: Optimizing Evaluation Through Machine Learning

Abstract Information: Vast amount of text data (e.g., interview transcripts, academic texts, open-ended survey responses) are common in evaluation. The rich text data are informative but challenging to synthesize to help evaluators and legislators understand significant factors that are helpful in making decisions and deriving actionable insights from evaluations. Hence, to generate insights from text data, Machine Learning Text classification algorithms assist in understanding existing correlations between related factors and educational outcomes. In this study, we explore data preprocessing techniques such as tokenization and vectorization to synthesize educational text data. We then evaluate the performance of machine learning classifiers such as random forest, and neural network. To demonstrate the classifiers' robustness and accuracy, we present the results of the validity measures (accuracy, precision, recall, and F1 score).