SCC: An approach to improve snippets comparison by classification
Gong Bihong · Journal of Tsinghua University(Science and Technology) · 2005
A snippet is a short document in which most occur only once, which makes conventional document comparison methods ineffective, such as the TF-IDF and cosine. An algorithm, SCC, was developed to solve the problem using classification method to expand the information of short snippets and then improve the snippet comparison. Two snippets are compared with the reference to obtain document sets similar to the snippets in large document sets, which are then used to expand the snippets to evaluate the snippet similarity. Experimental results show that the discounted cumulative gain DCG benchmark of relevance sets is improved significantly with the improved snippet comparison method being used.