DCU@Morpheme Extraction Task of FIRE-2012
Debasis Ganguly, Johannes Leveling, Gareth J. F. Jones · 2013
For the participation of Dublin City University (DCU) in the FIRE-2012 Morpheme Extraction Task (MET), we investigated rule based stemming approaches for Bengali and Hindi IR. The MET task itself is an attempt to obtain a fair and direct comparison between various stemming algorithms by measuring the relative gains in retrieval effectiveness as obtained with each stemming algorithm on the same dataset. In our stemming approach, linguistic knowledge was used to manually craft the rules for removing the commonly occurring plural suffixes for Hindi and Bengali. Additionally, rules for removing classifiers and case markers in Bengali were also formulated. Our rule-based stemming algorithms produced the best and the second-best retrieval effectiveness for Hindi and Bengali datasets respectively.