Error intercepted word screening method and system based on n-gram model

Abstract

The invention discloses an error intercepted word screening method and system based on an n-gram model, and relates to the technical field of network security. The method comprises the steps: acquiring audio translation text data intercepted based on an intercepted word under a specific label; processing the text data through the n-gram model, and screening out data which is not stored in the specific label from the text data as backspacing information; and determining a sentence containing the error intercepted word according to the backspacing information. The method is suitable for interception of forbidden words and sensitive words, especially for interception of the forbidden words and the sensitive words of audio translation text data, can quickly find out mistakenly intercepted sentences and mistakenly intercepted words, and can improve and optimize a forbidden word bank subsequently according to the obtained mistakenly intercepted words; therefore, the interception accuracy of the corresponding intercepted words and the overall interception accuracy are improved.

Description

technical field [0001] The invention relates to the technical field of network security, in particular to an n-gram model-based error blocking word screening method and system. Background technique [0002] The content on the Internet is increasing day by day, and the information often contains illegal and illegal information. Therefore, it is necessary to review and filter these contents to ensure a safe Internet environment and business needs. [0003] At present, the review method is usually to set up prohibited words and user-defined black / white thesaurus to block prohibited words and sensitive words. However, the existing interception methods only intercept the words themselves, and it is difficult to mine the semantics of the context, so the accuracy of the interception is low, especially for the data interception of speech-to-text, due to the existence of homonyms, words with similar pronunciation and dialects etc., leading to a further reduction in the accuracy of i...

AI Technical Summary

Problems solved by technology

However, the existing interception methods only intercept the words themselves, and it is difficult to mine the semantics of the context, so the accuracy of the interception is low, especially for the data interception of speech-to-text, due to the existence of homonyms, words with similar pronunciation and dialects etc., leading to a further reduction in the accuracy of interception

Images