Using the Divergence Framework for Randomness: CHiC 2013 Lab Report

For this first participation to the CHiC Lab, we focused on understanding the challenges of working with a collection of cultural heritage objects with short textual descriptions and on how to fine-tune a set of weighting models from the probability models based on Divergence From Randomness to perform uniformly in monolingual and multilingual scenarios. The official runs submitted used PL2 as the retrieval model and query expansion for four monolingual runs for English and Italian, and two multilingual runs against an English-Italian collection. Our best results were obtained in the unofficial runs using DLH13 with stemming and stopwords removal.