| 摘要 |
A method and a system for filtering harmful contents are provided to filter adult or AD(ADvertisement) contents with high accuracy by using a 3-tier harmful contents blocking step, and reduce accidents continuously and increase reliability of a filter by increasing the number of contents stored in an adult/AD contents database. A harmful contents database(150) stores contents judged as harmful contents. An image register(140) stores the contents judged as normal contents. A classifier(110) classifies inputted contents into one category. A filter(120) calculates probability for the harmful contents by extracting text included in the inputted contents, and analyzing an area of a person or an object with the filter assigned to the category. A matching engine(130) calculates similarity by comparing the harmful contents with the inputted contents. The filter transfers the inputted contents to the harmful contents database when the similarity is higher than a first threshold, transfers the inputted contents to the image register when the similarity is lower than a second threshold, and transfers the inputted contents to the matching engine when the similarity is higher than the second threshold and lower than the first threshold.
|
