A model for evaluating the quality of user-created documents

Abstract

In this paper, we propose a model for evaluating the quality of general user-created documents. The model is based on supervised classification approach, in which output scores are considered as quality of given document. In order to utilize both textual and non-textual attributes of documents, we incorporated a number of objectively measurable, real-valued features selected upon predefined criteria for quality. Experiments on two datasets of real world documents show that textual features are stable indicators for evaluating documents' quality. Some features are inferred to be effective for general kinds of documents. © 2008 Springer-Verlag Berlin Heidelberg.