This paper presents a knowledge-basedmethod for measuring the semanticsimilarityof texts. While there is a largebody of previous work focused on findingthe semantic similarity of conceptsand words, the application of these wordorientedmethods to text similarity has notbeen yet explored. In this paper, we introducea method that combines wordto-word similarity metrics into a text-totextmetric, and we show that this methodoutperforms the traditional text similaritymetrics based on lexical matching.
展开▼