Automatic address correction method for man-machine fusion customer service system

The invention discloses an automatic address correction method for a man-machine fusion customer service system, which comprises the following steps of: converting address voice into an address text through an ASR model, processing the address text through an NLP model to form a corrected address text, obtaining structured data of address information, finally, inputting a POI query interface of a third-party map supplier to obtain final address information; carrying out pinyin coding on each input sample in the data set used for NLP model training, confusing the input samples according to common voice errors, forming confusion sets of the input samples, taking the input samples and the confusion sets of the input samples as amplification input samples, and training the NLP model. Aiming at common voice errors existing in address voice, pinyin codes of input samples and confusion sets of the pinyin codes are utilized to amplify training data of an NLP model, the method has the capability of distinguishing the