Method for realizing language recognition based on deep learning

Abstract

The invention discloses a method for realizing language recognition based on deep learning, which comprises the following steps: after obtaining a call recording file, generating a language text dataset through AliCloud ASR and a language recognition interface; performing language text noise reduction processing on the feedback recognition result; language texts under the categories are extractedfor recognition, language judgment is conducted on the categories, and the training corpus manufacturing procedure is completed; words of the training corpus are mapped into index representation; constructing a vocabulary-index mapping table, constructing a label-index mapping table for language labels, reading word vectors from a pre-trained word vector model, inputting the word vectors as initialization values into the model, performing numeralization on language texts and language labels through the mapping table to obtain index expressions, filling the index expressions with fixed lengths, and submitting the index expressions to a deep learning classifier for training; and the deep learning classifier analyzes and predicts the language text to be tested, and finds out the language category with the highest probability. The method can reduce the manual rehearing pressure, saves the manpower, is high in efficiency, is automatic, and is high in accuracy.

Description

technical field [0001] The invention relates to the field of telecommunications, in particular to a method for realizing language recognition based on deep learning. Background technique [0002] At present, there is a lack of language data for customer service hotlines, but user attribute characteristics can be mined in call content, such as language type, address information, business demand content, etc. In the follow-up business analysis, it is necessary to mine the changes in the business needs of each user group from the basic attributes of users, improve the complaint monitoring system of each user group, and provide favorable data support for subsequent refined user operation and maintenance. In the absence of basic user attribute indicators (language type), it is necessary to manually label the recording data. However, since 20,000 to 30,000 users call the customer hotline every day to inquire about the business situation, if the telecom operator needs to classify ...

AI Technical Summary

Problems solved by technology

However, since 20,000 to 30,000 users call the customer hotline every day to inquire about the business situation, if the telecom operator needs to classify the language of the customer service hotline service, it will take a lot of manpower to listen to and mark the call recordings every day. Manual review and language labeling will consume a lot of manpower and time

Images