Automatic wrongly written character correcting method in search engine and server

Abstract

The invention provides an automatic wrongly written character correcting method in a search engine and a server. The wrongly written character in the text can be corrected more effectively; each character is mapped to a high-order space by deeply learning a model and high-dimensionally vectoring, the relation between the characters is represented through the high-dimension vector, and the context information of the character and the effect thereof in the sentence are used for recognizing whether the character is the wrongly written character. According to the method provided by the invention, a one-to-one corresponding relation between the wrongly written characters and the correct characters does not need to be constructed by consuming a range of costs, but only a proper wrongly written training correction model needs to be constructed to learn the features of the wrongly written character. The wrongly written character in the sentence is recognized and corrected by considering the semantic and grammar of the sentence, the word characteristic and the context information of the word by the technical scheme adopted by the invention, the recognized wrongly written character is not only the homophonous character or characters with the similar form, and according to the technical scheme, the wrongly written words of other types can be further recognized and corrected, and the correction efficiency of the wrongly written character is greatly improved.

Description

technical field [0001] The invention relates to the field of data correction methods, and more particularly, to an automatic correction method and server for typos in a search engine. Background technique [0002] The main technique in the text proofing process is to correct typos. Usually, the text proofreading process basically adopts two methods (manual checking proofreading and proofreading based on the typo dictionary), the most important of which is the typo dictionary proofreading, by constructing a thesaurus corresponding to the wrong words and the correct words. The Chinese patent "CN1116343A Chinese typo automatic correction method and device" provides a dictionary-based typo correction method. The invention constructs a dictionary of typos by finding a large number of words based on similar glyphs, pronunciations or input codes as word pairs, and then uses a scoring model to score the corresponding words, and finally selects the correct word from the dictionary a...

AI Technical Summary

Problems solved by technology

The present invention well solves the shortcomings of the typo correction system under the traditional method

Images