Mining of relevant and informative posts from text forums

Abstract

In the modern world, the competitive advantage for every person is the possibility to obtain the information in a fast and comfortable way. Web forums occupy a significant place among the sources of information. It is a good place to gain professionally significant knowledge on different topics. However, sometimes it is not easy to identify the places on the forum, which contains useful information corresponding user demands. In this paper we consider the problem of automatic forum text summarization and describe the methods, which can help to solve it. We study the difference between relevance-oriented and useful-oriented query types. We will describe our dataset, that contains over 4000 of marked posts from web forums about various subject domains. The posts were marked by experts, by estimating them on a scale from 0 to 5 for selected query types. The results of our study can provide background for creation informational retrieval applications that will decrease the time of user’s searching and increase the quality of search results.