Multiattribute context analysis to "unclarity" sensibility and applications to malicious information judgment on the Web

互联网已经在世界范围内广泛传播。另一方面，公告牌上存在着大量的滥用、危险和犯罪等恶意信息，这已经成为一个社会问题。由于手工提取恶意信息是不可能的，因此需要一种自动检测恶意信息的方法，在本研究中，大规模地收集自然清晰的表情，并利用这些表情构建的检测规则来自动提取恶意信息。实验结果表明，自动检测的成功率在70%~80%之间。

Internet has spread widely in the world. On the other hand, there is much malicious information such as abuse, danger and crime in bulletin boards and this becomes a social problem. An automatic method of detecting the malicious information should be required, because it is impossible to extract it by hand.In this research, "natural and clear expressions" is collected on a large scale and the detecting rules built from these expressions is utilized to extract malicious information automatically. According to experimental results, the achievement of automatic detection is from 70% to 80%.