Search

In this paper, we have developed a probabilistic approach using PLSA for the discovery and analysis of contextual keyword relevance based on the distribution of keywords across a training text corpus. We have shown experimentally, the flexibility of this approach in classifying keywords into different domains based on their context. We have developed a prototype system that allows us to project keyword queries on the loaded PLSA model and returns keywords that are closely correlated. The keyword query is vectorized using the PLSA model in the reduce aspect space and correlation is derived by calculating a dot product. We also discuss the parameters that control PLSA performance including a) number of aspects, b) number of EM iterations c) weighting functions on TDM (pre-weighting). We have estimated the quality through computation of precision-recall scores. We have presented our experiments on PLSA application towards document classification.

In this paper, we show that the efficiency of LSA is significantly controlled by the choice of weighting algorithm applied. These weighting algorithms allocate relative importance to the document attributes (e.g. Keywords) based on their occurrences in the corpus. Effects of different weighting algorithms will be the central point of this paper. We experimented with various weighting algorithms to evaluate and study their effects as measured by precision and recall values. Our experiments include weighting function application on TDM (Pre-Weighting) in order to increase or decrease the relative importance of words based on their occurrence. We also evaluated the application of weighting functions on the projected query (post-weighting). Post-weighted keyword queries were projected on an LSA model built on pre-weighted TDM to obtain closely correlated keywords or a document (keyword collection). We have developed a prototype IR query projection tool which projects keyword queries on the LSA model to retrieve relevant keywords with a floating-point score.