Method and system for identifying and verifying inflected words

Abstract

The invention discloses a recognizing and verifying method and system for an anagram. The method has the following beneficial effects: an anagram library can be expanded through the pronunciation and character pattern expansion, and thus the quantity of the anagrams in the anagram library can be increased, the quality can be improved, and the misjudgement probability can be decreased; a training sample is used for training context probability, thus the misjudgement probability in semantic verification of the anagram can be further decreased, and moreover, the accuracy can be improved; the training sample is updated based on the verification result, so that the automatic updating performance of algorithm can be improved, and as a result, a concept base for the semantic verification can be expanded; as the accumulation of the verification results, the misjudgement probability continuously decreases. The recognizing and verifying system comprises an acquiring unit, an anagrammatizing training unit, a recognizing unit and a semantic verification unit. The system has the same beneficial effects as the method.

Description

Technical field [0001] The invention relates to the field of machine recognition of deformed words, in particular to a method and system for identifying and verifying deformed words. Background technique [0002] We often see deformed sensitive words in various platforms such as post bars, forums, and news media. The way of thinking of the human brain allows us to find these deformed words very naturally, because these deformed words are "abnormal" parts in the sentence. This "abnormal" feeling will focus our attention on this area, and then Gradually find complete inflections. When the machine directly faces these deformed words (including intermixed special symbols, homophone transformation, form near transformation, simplified and traditional conversion, radical splitting, etc.), it appears to be a little weak. The recognition of deformed words is an important problem in solving Chinese spam content filtering. . [0003] At present, in the Chinese patent application with appl...

AI Technical Summary

Problems solved by technology

[0005] The disadvantage of the existing inflected word recognition technology is that the system and method are based on a fixed inflected thesaurus and concept library, and the number and quality of inflected words and training samples have relatively large limitations, which is prone to misjudgment ; The automatic update performance of the algorithm is poor, and it does not have the ability to expand the variable word thesaurus and concept library

Images