A Comparative Study on Ontological Method with Deep Learning
- 주제(키워드) Sentiment Analysis , Context-Aware Technologies , Ontology-Based Methods , Deep Learning Integration , Semantic Understanding , Complex Context Handling
- 주제(DDC) 006.31
- 발행기관 아주대학교 일반대학원
- 지도교수 Seok-Won Lee
- 발행년도 2025
- 학위수여년월 2025. 2
- 학위명 박사
- 학과 및 전공 일반대학원 인공지능학과
- 실제URI http://www.dcollection.net/handler/ajou/000000034633
- 본문언어 영어
- 저작권 아주대학교 논문은 저작권에 의해 보호받습니다.
초록/요약
In the rapidly advancing domain of artificial intelligence, sentiment analysis and context-aware technologies have emerged as critical research areas. Traditional emotion analysis models exhibit notable limitations in processing context-dependent emotions, sarcasm, and nuanced emotional expressions, especially when analyzing sentiment trends in social media comments or literary works. These models frequently fail to accurately capture intricate emotions influenced by contextual shifts, leading to significant deviations in analysis outcomes. Current research suggests that deep learning models alone are inadequate for addressing such complex scenarios. Conversely, while ontology-based context-aware and activity recognition technologies hold promise for intelligent system applications, they encounter substantial performance challenges in handling complex contexts and abnormal activities in domains such as smart home environments and intelligent traffic monitoring. Although deep learning approaches excel in automatic feature learning, they inherently lack robust semantic logic comprehension and uncertainty management capabilities. Ontology methods, while adept at constructing structured knowledge systems that clearly delineate conceptual relationships, exhibit certain limitations in data-driven adaptability. Our study introduces a novel approach by deeply integrating ontology- based methods with deep learning models to address these challenges. Through meticulously designed explicit and implicit topic identification strategies, the proposed framework effectively mitigates context interference and precisely captures core emotional cues within texts. By leveraging multiple advanced techniques, an extended sentiment lexicon and efficient identification methods are developed to enhance the model's comprehension of emotional semantics. The framework's performance is rigorously evaluated using a restaurant review dataset as a primary testbed. Experimental results demonstrate substantial improvements in sentiment analysis accuracy, marking a significant advancement in affective computing. Moreover, the optimized framework exhibits superior performance in multi-scenario intelligent system applications. Extensive validation shows marked improvements in recognition accuracy and reduced processing time compared to traditional methods, particularly in complex scenarios and uncertain data processing tasks. A comprehensive comparison between ontology and deep learning methods reveals that ontology approaches offer unique advantages in handling complex contextual information, providing more reliable and efficient decision support for intelligent systems. The integration of ontology methods with deep learning introduces domain knowledge on a data- driven basis, enhancing model generalization and robustness while supporting more complex inference tasks. This ontology-enhanced deep learning approach provides a new direction for developing more intelligent and transparent artificial intelligence systems, offering broad application prospects and profound research significance in advancing intelligent systems to new levels of sophistication. Keywords: Sentiment Analysis, Context-Aware Technologies, Ontology-Based Methods, Deep Learning Integration, Semantic Understanding, Complex Context Handling
more목차
CHAPTER 1. INTRODUCTION 1
1.1. Background and Motivation 1
1.2. Research Contributions 6
1.3. Organization of Dissertation 10
CHAPTER 2. RELATED WORKS 12
2.1. Integration of ontology and deep learning 12
2.2. Comparative Study of Ontology and Deep Learning 19
CHAPTER 3. INTEGRATING ONTOLOGY-BASED APPROACHES WITH DEEP LEARNING MODELS 30
3.1. Problem Statement 30
3.2. Motivation 33
3.3. Proposed Approach and Experiments 36
3.4. Contribution 46
CHAPTER 4. A COMPARATIVE STUDY ON ONTOLOGICAL METHOD WITH DEEP LEARNING 48
4.1. Problem Statement 48
4.2. Motivation 52
4.3. Proposed Approach and Experiments 55
4.4. Contribution 73
CHAPTER 5. DISCUSSION 75
5.1. The integration of ontology methodology and deep learning 75
5.2. Theoretical Innovation of Ontology Methodology and Deep Learning 77
5.3. The Practical Application Value of Ontology Methodology and Deep Learning 78
5.4. The Limitations of the Ontological Methodology and Deep Learning 80
CHAPTER 6. CONCLUSIONS 82
6.1. Conclusions 82
6.2. The Practical Application Value 84
6.3. Future work 85
REFERENCES 88
