A Deep Semantic Comprehension-Based Automatic Conversation Model Based on Autoencoder-Enhanced Transformer

• Published in: IEEE access Vol. 13; pp. 97975 - 97986
• Main Authors: Liang, Shouyu, Zhao, Bimei, Li, Cheng, Chen, Yuanfeng, Ren, Zhengguo, Ao, Bang
• Format: Journal Article
• Language: English
• Published: IEEE 2025
• Subjects: autoencoder; Autoencoders; automatic conversation; Computational modeling; Decoding; Feature extraction; Natural language processing; Neural networks; Oral communication; Semantic comprehension; Semantics; transformer; Transformers; Vectors
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2025.3572384

With the continuous advancement of artificial intelligence and machine learning technology, automatic dialogue systems have become an indispensable part of people's daily lives. However, the complexity of semantic understanding makes it difficult for systems to accurately capture user intent and contextual information. Dialogue systems often appear stiff and disjointed when generating natural language, lacking the fluency and naturalness of human dialogue. Therefore, a model is needed to deal with these problems. This article proposes an automatic conversation model based on deep semantic understanding, which aims to improve the performance of natural language processing systems in dialogue scenarios and achieve more efficient and accurate semantic understanding and generation. The study introduced AutoEncoder into the Transformer structure and designed a multi-level autoencoder network to achieve deeper semantic understanding of input text. The AutoEncoder based on deep semantic algorithm was developed and a new automatic session awareness module was created. The study used the DailyDialog and Ubuntu dialogue datasets, and through comparative experiments and qualitative analysis, we validated the effectiveness and superiority of the proposed model in dialogue scenarios. Our model has achieved significant improvements in semantic understanding and generation, demonstrating its potential application prospects in practical dialogue systems. The automatic conversation model based on deep semantic understanding and the Autoencoder enhanced Transformer proposed in this article provide a new approach and method for improving the performance and efficiency of dialogue systems.