[Paper Review] Query Optimization Using Genetic Algorithms in the Vector Space Model

Eman Al Mashagba, Feras Al Mashagba+1 more|arXiv (Cornell University)|Nov 30, 2011

Metaheuristic Optimization Algorithms Research14 references13 citations

TL;DR

This study applies genetic algorithms (GA) to optimize query retrieval in Arabic text collections using the Vector Space Model (VSM), evaluating multiple similarity measures (Dice, Inner Product) and GA configurations. The best-performing system used Inner Product as the fitness function with a one-point crossover, significantly improving retrieval effectiveness over other GA strategies.

ABSTRACT

In information retrieval research; Genetic Algorithms (GA) can be used to find global solutions in many difficult problems. This study used different similarity measures (Dice, Inner Product) in the VSM, for each similarity measure we compared ten different GA approaches based on different fitness functions, different mutations and different crossover strategies to find the best strategy and fitness function that can be used when the data collection is the Arabic language. Our results shows that the different GA approaches have differences in their results, the best IR system found is the one that uses the Inner Product similarity as a fitness with one-point crossover operator.

Motivation & Objective

• To enhance query retrieval effectiveness in Arabic text collections using genetic algorithms within the Vector Space Model framework.
• To evaluate the impact of different similarity measures—Dice and Inner Product—on retrieval performance in Arabic IR systems.
• To identify the optimal combination of GA parameters, including fitness functions, crossover strategies, and mutation operators, for Arabic language data.
• To determine whether genetic algorithms can effectively optimize query weighting and document ranking in large-scale Arabic information retrieval tasks.

Proposed method

• Employed the Vector Space Model (VSM) to represent documents and queries as vectors in a high-dimensional space using Arabic text collections.
• Implemented ten distinct genetic algorithm (GA) configurations, varying fitness functions, crossover operators (including one-point, two-point, and uniform), and mutation strategies.
• Used the Inner Product and Dice coefficient as similarity measures to compute fitness values during GA evolution.
• Evolved query weight vectors through selection, crossover, and mutation to maximize relevance scores based on the chosen fitness function.
• Evaluated system performance using standard IR metrics, comparing results across different GA parameter combinations.
• Selected the best-performing configuration based on retrieval effectiveness, focusing on precision and recall in Arabic document retrieval.

Experimental results

Research questions

• RQ1 Which similarity measure—Dice or Inner Product—yields better retrieval performance when used as a fitness function in a genetic algorithm for Arabic IR?
• RQ2 How do different crossover strategies (e.g., one-point, two-point, uniform) affect the optimization of query weights in the VSM for Arabic text?
• RQ3 What is the optimal combination of fitness function, mutation strategy, and crossover operator for improving retrieval effectiveness in Arabic information retrieval?
• RQ4 Can genetic algorithms effectively optimize query representation in the vector space model for Arabic documents, outperforming baseline methods?

Key findings

• The Inner Product similarity measure produced superior retrieval results when used as the fitness function in the genetic algorithm.
• The one-point crossover operator consistently outperformed other crossover strategies in terms of retrieval effectiveness.
• The combination of Inner Product as the fitness function and one-point crossover yielded the highest precision and recall scores among all evaluated GA configurations.
• Different GA parameter combinations led to significant variations in retrieval performance, indicating that parameter selection is critical for optimization success.
• The proposed GA-based approach improved retrieval effectiveness over standard VSM methods, particularly in the context of Arabic language text collections.
• Mutation strategy had a lesser impact compared to fitness function and crossover type, but still contributed to convergence and diversity in the search space.