In our paper Term frequency normalisation tuning for BM25 and DFR model in the 27th European Conference on Information Retrieval (ECIR 2005), we described our methodology for the collection-dependent problem of term frequency normalisation
for BM25 and DFR models.
Moreover, we have published the paper Inferring Query Performance Using Pre-retrieval Predictors in the Proceedings of the Eleventh Symposium on String Processing and Information Retrieval (SPIRE 2004). This paper studies a set of query features that were useful in predicting query performance. An extended journal version is accepted in the Special Issue of Information Systems
for the SPIRE2004.
We have extended our study to the smoothing technique in language modelling. In our paper A Study of Dirichlet Priors for Term Frequency Normalisation, which is published in the 28th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2005), we applied the Dirichlet Priors for the term frequency normalisation of the Divergence From Randomness models, as well as the BM25 probabilistic model.
In addition, in our submitted paper to the ACM Transactions on Information Systems (TOIS), we proposed a theoretically-driven automatic parameter setting approach to the term frequency normalisation. This approach provides a better understanding of term frequency normalisation. We applied the propsoed theorectically-driven parameter setting approach in our participation in the 14th Text REtrieval Conference (TREC 2005).
The publications of the Smooth project can be found at the Publications section of this website.