Akademik Veri Yönetim Sistemi
BOBCATSSS 2025: Artificial Intelligence in Library and Information Science: Exploring the Intersection, İstanbul, Türkiye, 21 - 23 Ocak 2025, ss.1-12, (Tam Metin Bildiri)
Traditional information retrieval systems have several problems such as limited understanding of user intentions, lack of flexibility in natural language queries, static indexing, scalability, relevance and ranking limitations. Artificial Intelligence (AI) is now integrated into various aspects of life, from search engines to archive systems. If the precision and relevance of the results of traditional information retrieval systems are increasing day by day, it is thanks to the integration of AI. However, given that AI is ultimately composed of algorithms with a certain bias, it is anticipated that there may be many types of data or information that it misses or omits. A Knowledge Graph (KG) approach to be integrated into AI-based search engines could provide more data and therefore more relevance and precision. While search engines with AI are largely capable of solving these problems, integrating KGs would add additional features such as semantic search, annotation, context-sensitive retrieval, complex query understanding, etc., and could further improve existing high scores. The function of the KGs approach here is to be fully inclusive to further improve contextual retrieval. It would be integrated into classical AI phenomena such as Machine Learning (ML) and Natural Language Processing (NLP), thus contributing to more efficient results through its inclusion function. Compared to alternatives such as relational or NoSQL databases found in various applications, the KGs based approaches are known to be particularly useful in abstraction. The main research question in this study is how KGs can improve relevance and precision scores in information retrieval. The paper initially reviews the literature on KGs and provides a basic definition of it, then evaluates the functioning of AI-driven search engines in terms of relevance and precision scores and discusses how the proposed approach can be integrated into these processes. In the conclusion, inferences are drawn in terms of relevance metrics (precision, sensitivity, F1) and semantic metrics (entity disambiguation accuracy, information completeness), which are the main metrics that can be used to evaluate the KGs approach. The study provides an insight into different aspects of the use of AI in information retrieval and thus knowledge discovery, and provides an updated perspective on both traditional and new AI-driven knowledge discovery applications, as well as new future prospects