This paper presents two ACO algorithms for information retrieval. Unlike existing works, the proposed algorithms address the problem for large scale data sets. The algorithms and a classical information retrieval method have been implemented for comparison purposes. Experimentations have been conducted on smart collections and random benchmarks. Numerical results show that for small collections of documents the classical approach is faster than ACO algorithms whereas for large scale data, ACO is much more interesting in terms of runtime and performs as well as the exhaustive search. The novel outcome of this study consists in determining the frontier in terms of collection size from which ACO outperforms the classical information retrieval algorithm especially from the runtime point of view.