Enhancing Personalized Explainable Recommendations with Transformer Architecture and Feature Handling DOI Open Access
Ming-Yen Lin,

I-Chen Hsieh,

Sue-Chen Hsush

et al.

Electronics, Journal Year: 2025, Volume and Issue: 14(5), P. 998 - 998

Published: Feb. 28, 2025

The advancement of explainable recommendations aims to improve the quality textual explanations for recommendations. Traditional methods primarily used Recurrent Neural Networks (RNNs) or their variants generate personalized explanations. However, recent research has focused on leveraging Transformer architectures enhance by extracting user reviews and incorporating features from interacted items. Nevertheless, previous studies have failed fully exploit relationship between ratings more In this paper, we propose a novel model named EPER (Enhanced Personalization Explainable Recommendation), which considers reviews, ratings, feature words, item titles high-quality employs masking mechanism prevent interference rating prediction explanation generation. Moreover, an innovative feature-handling method manage missing interaction in existing models. Experimental results public datasets demonstrate that generally outperforms other well-known methods, including NETE, PETER+, MMCT. Compared with MMCT, improves (ROUGE metric) 3.27%, personalization (FMR 6.82%, (MSE 1.2% Amazon Clothing dataset. Overall, provides recommendation match exceed best demonstrating its potential practical applications.

Language: Английский

Enhancing Personalized Explainable Recommendations with Transformer Architecture and Feature Handling DOI Open Access
Ming-Yen Lin,

I-Chen Hsieh,

Sue-Chen Hsush

et al.

Electronics, Journal Year: 2025, Volume and Issue: 14(5), P. 998 - 998

Published: Feb. 28, 2025

The advancement of explainable recommendations aims to improve the quality textual explanations for recommendations. Traditional methods primarily used Recurrent Neural Networks (RNNs) or their variants generate personalized explanations. However, recent research has focused on leveraging Transformer architectures enhance by extracting user reviews and incorporating features from interacted items. Nevertheless, previous studies have failed fully exploit relationship between ratings more In this paper, we propose a novel model named EPER (Enhanced Personalization Explainable Recommendation), which considers reviews, ratings, feature words, item titles high-quality employs masking mechanism prevent interference rating prediction explanation generation. Moreover, an innovative feature-handling method manage missing interaction in existing models. Experimental results public datasets demonstrate that generally outperforms other well-known methods, including NETE, PETER+, MMCT. Compared with MMCT, improves (ROUGE metric) 3.27%, personalization (FMR 6.82%, (MSE 1.2% Amazon Clothing dataset. Overall, provides recommendation match exceed best demonstrating its potential practical applications.

Language: Английский

Citations

0