Patrick Moran, College of Charleston


Feature Extraction from Textual Datasets


Abstract: A new method is presented for topic extraction from sets of textual documents. This method takes into account word frequencies, semantic similarity of words, and proximity of words in the dataset. Existing techniques such as a nonnegative matrix factorization and graph clustering algorithms are employed. This method was developed and tested on sets of camera reviews, and results will be presented.

Advisor: Prof. Carl Meyer (North Carolina State University)