This technology is a multilingual URL filtering device consisting of a local ontology containing a multilingual vocabulary dictionary, a whitelist, and a blacklist; a web filtering system that calculates URL evaluation scores to determine whether to allow or block access; and a RESTful web service module that relays requests and responses. For URLs that cannot be determined by the local ontology, the class is decided using the YAGO ontology, which includes WordNet and Wiki-based ontologies.
While URL-based filtering is fast, it often lacks accuracy because URLs do not always reflect actual web page content. Conversely, content-based filtering offers high accuracy but suffers from processing delays and high computational costs. Furthermore, most existing AI-based URL filtering systems are limited to English URLs, failing to handle multilingual URLs or resulting in excessive blocking.
This technology retrieves titles, keywords, and description metadata from the head section of a web page for newly encountered URLs, determining whether to block access based on a final evaluation score that combines the URL, title, and metadata scores. Once processed, the URL and its evaluation score are stored in the local ontology, allowing for instant determination upon re-request. Applicable to school and home parental control services, corporate security gateways, and ISP-level content filtering, it accurately filters out harmful sites in various languages without latency.
N/A
