PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2166 - e2166

Published: July 3, 2024

Amid the wave of globalization, phenomenon cultural amalgamation has surged in frequency, bringing to fore heightened prominence challenges inherent cross-cultural communication. To address these challenges, contemporary research shifted its focus human–computer dialogue. Especially educational paradigm dialogue, analysing emotion recognition user dialogues is particularly important. Accurately identify and understand users’ emotional tendencies efficiency experience interaction play. This study aims improve capability language It proposes a hybrid model (BCBA) based on bidirectional encoder representations from transformers (BERT), convolutional neural networks (CNN), gated recurrent units (BiGRU), attention mechanism. leverages BERT extract semantic syntactic features text. Simultaneously, it integrates CNN BiGRU delve deeper into textual features, enhancing model’s proficiency nuanced sentiment recognition. Furthermore, by introducing mechanism, can assign different weights words their tendencies. enables prioritize with discernible inclinations for more precise analysis. The BCBA achieved remarkable results classification tasks through experimental validation two datasets. significantly improved both accuracy F1 scores, an average 0.84 score 0.8. confusion matrix analysis reveals minimal error rate this model. Additionally, as number iterations increases, recall stabilizes at approximately 0.7. accomplishment demonstrates robust capabilities understanding showcases advantages handling characteristics expressions within context. proposed provides effective technical support which great significance building intelligent user-friendly systems. In future, we will continue optimize structure, complex emotions cross-lingual recognition, explore applying practical scenarios further promote development application dialogue technology.

Language: Английский