TITCH: Attribute selection based on discrimination power and frequency

Philipp Spanger, Takahiro Kurosawa, Takenobu Tokunaga · Institutional Repositories DataBase (IRDB) · 1995

We provide a simple base algorithm for the given task of attribute selection as well as two improvements to this algorithm. We report on the results of their implementation and provide an error analysis. We then report some observations and conclusions about the human attribute selection process by comparing the output of our algorithms and the attributes selected by humans. Finally, we add some observations on the difficulty of the attribute selection task. Note: We utilised both training and development data for the development and evaluation of our algorithm.

Read the paper · More papers on PaperTik