Emotion Vocabulary Recognition in English Writing Based on the B-BLLC-CL Algorithm

• Li Ren,
• Shaidatul Akma Adi Kasuma,

Vol. 19, No. 9, September 30, 2025
10.3837/tiis.2025.09.001, Download Paper (Free):

• Emotion Vocabulary Recognition
• B-BLLC-CL Algorithm
• Text sentiment analysis
• Deep Learning
• Machine Learning

Abstract

With the advent of the digital age, the fields of education and academia increasingly rely on digital tools to analyze and assess student essays and academic writing. These texts not only convey information but also reflect the authors' emotional tendencies and attitudes. This study aims to explore and enhance sentiment analysis technology, particularly in the emotional recognition of student essays, to help educators more accurately assess and guide students' abilities in emotional expression. We utilize a BERT-based entity recognition algorithm that significantly enhances the model's text representation capabilities by weighing and merging features vectors from different BERT layers. Additionally, we propose a text emotion recognition model named B-BLLC-CL. This model enhances the emotional characteristics of texts by incorporating emotion label information and uses a bidirectional long short-term memory network (Bi-LSTM) to capture the contextual knowledge of the text. By learning label features through a label encoder and constructing a relevancy matrix between labels and word representations using a label attention mechanism, the model explicitly models the relationship between emotion labels and texts. Experimental results on our custom-built writing emotion detection task dataset show that the B-BLLC-CL model improves the evaluation metrics of label ranking average precision, accuracy, micro-average F1 score, and macro-average F1 score by 1.57%, 1.95%, 1.91%, and 3.4% respectively, compared to the baseline model, and reduces the Hamming loss by 0.49%.

Statistics

Cite this article