Filtering product reviews from web search results

This study seeks to develop an automatic method to identify product reviews on the Web using the snippets (summary information) returned by search engines. Determining whether a snippet is a review or non-review is a challenging task, since the snippet usually does not contain many useful features for identifying review documents. Firstly we applied a common machine learning technique, SVM (Support Vector Machine), to investigate which features of snippets are useful for the classification. Then we employed a heuristic approach utilizing domain knowledge and found that the heuristic approach performs equally well as the machine learning approach. A hybrid approach which combines the machine learning technique and domain knowledge performs slightly better than the machine learning approach alone.