Evaluation of baseline information retrieval for Polish open-domain Question Answering system

Michał Marcińczuk, Adam Radziszewski, Maciej Piasecki, Dominik Piasecki, Marcin Ptak · 2013

We report on our efforts aimed at building an Open Domain Question Answering system for Polish. Our contribution is twofold: we gathered a set of question–answer pairs from various Polish sources and we performed an empirical evaluation of two re-ranking methods. The gathered collection contains factoid, list, non-factoid and yes-no questions, which makes a challenging material for experiments. We show that using two re-ranking methods based on term proximity allows to obtain significant improvement on simple information retrieval baseline. The improvement is observed as finding more answer-bearing documents among the top n search results. 1

Read the paper · More papers on PaperTik