A Multi-Objective Evolutionary Algorithm Approach to Learn Disjunctive and Conjunctive Topical Queries

CECCHINI ROCIO LUJÁN; LORENZETTI CARLOS MARTÍN; MAGUITMAN ANA GABRIELA

Mar del Plata, Buenos Aires

Congreso; Jornadas Argentinas de Informatica e Investigacion Operativa; 2009

Sociedad Argentina de Informática

Topical search refers to the process of formulating queries that reflect a thematic context.  A combination of machine learning and information retrieval techniques can be applied to automate this process. In this work we propose the application of single- and multi-objective evolutionary algorithms to automatically evolve a population of topical queries. We report on the results of different strategies that attempt to evolve conjunctive and disjunctive queries. In our evaluations we observe that disjunctive queries have the potential to achieve better retrieval performance than conjunctive queries. In addition, our analysis  reveals the limitations of the single-objective approach and highlights the advantages of applying multi-objective evolutionary algorithms for the problem at hand.