A cooperative co-evolutionary genetic algorithm for query recommendation

Search engines often recommend a few queries related to the users' original query to help them find the content they are searching. Since extracting the user's intent from the query is a very challenging task due to its short length and ambiguity, it is tough to build a good Query Recommendation system. In this work, we have proposed a Query Recommendation system using a Multi-objective Cooperative Co-evolutionary Genetic Algorithm (QRMOCCGA) to address the problem. First, we have decomposed the entire problem into two sub-problems. These sub-problems optimise two objective functions, created using users' search behaviours and various string-based similarity methods. QRMOCCGA uses separate sub-populations to solve the sub-problems simultaneously. It finds complete Pareto-optimal solutions by assembling the relevant members from the two sub-populations which have been collaboratively co-evolved. QRMOCCGA also maintains diversity in the population. We perform extensive experiments to benchmark our proposed algorithm against several popular algorithms on a large-scale search log extracted from a commercial search engine.

Automated summary

This paper proposes a Query Recommendation system using a Multi-objective Cooperative Co-evolutionary Genetic Algorithm (QRMOCCGA) to address the challenges of understanding user intent and providing query recommendations.