MOSTER: A Novel Truth Discovery Method for Multiple Conflicting Information

With the rapid development of database and web technology, the way data organized and presented is becoming increasingly complicated while data sources are also intermingled with inaccurate information. Therefore, studies in truth discovery becomes overwhelmingly significant for it is critical for netizens to identify sources of high quality as well as to select the most accurate information from vast amount of data available. However, previous works mainly focus on a single property rather than all properties, and ignore the different characteristics of them, thus leading to unexpected deviations. In this paper, we propose a Multi-prOperty-cluSTERing-based method, abbreviated MOSTER, in order to search for the most reliable source and identify the truth. Compared with conventional methods, experiments using our three-step iterative approach can achieve a high accuracy both on weather data and people profile data, indicating a great advancement in truth discovery studies.