Overview of Boolean Search Queries Optimization in Information Retrieval Systems

Generating Boolean Search queries for patent retrieval integrates Machine Learning, Sophisticated computational approaches and informational retrieval techniques. With the goal to improve patent searches and refine queries, this paper determines recent improvements in the Boolean logic, optimization techniques, and AI-driven methods. Innovative approaches like SetBERT demonstrate that fine-tuning the neural models can improve the retrieval performance of Boolean queries like AND, OR, and NOT, which are necessary for patent search. By inversed-contrastive loss, SetBERT gives great recall performance and outperforms traditional BERT-based methods, with applications in patent information retrieval. Similarly, reinforcement learning-based pipelines in systematic literature reviews (SLRs) illustrate that an approach wherein query refinements are suggested based on real-time feedback shows the possibility of being extended to the optimization of Boolean queries for patent datasets. Optimization techniques are very important in patent searching in relation to both scalability and complexity. Thus, Quantum Approximate Optimization Algorithm (QAOA) offers scalable solutions to the problem of multi-query optimization for more effectively searching across large-sized patent databases. Advanced indexing methods such as BE-Tree and the graph-based elemental indexes enable efficient processing of high-dimensional Boolean queries, which is critical for exploration of nuanced semantic relationships in patent datasets. This would be very ideal for such applications as patent retrieval, where large data sets can be computationally intensive.