Method, device, electronic device and storage medium for predicting pronunciation of words

Abstract

The present application provides a method, device, electronic equipment and storage medium for predicting the pronunciation of a word. When predicting the pronunciation of a target word, first obtain each character vector of the target word, and then input the word pronunciation prediction model in turn to obtain the target word Predict pronunciation. Among them, the word pronunciation prediction model is obtained by using the corresponding relationship between multiple words and pronunciation factor sequences to train the deep neural network model in advance. The deep neural network model includes: an input layer, a recurrent neural network with at least two hidden layers, and output layer. This application improves the word pronunciation prediction model in the related art. Specifically, an improved LSTM neuron is used in the hidden layer. Compared with the conventional LSTM neuron, the improved LSTM neuron performs the input word vector The calculation speed can be significantly improved during abstract learning, therefore, when the pronunciation of words needs to be predicted, the method provided by this application can significantly improve the prediction efficiency.

Description

Technical field [0001] The present invention relates to the field of language processing technology, specifically to a method for predicting the pronunciation of words, apparatus, electronic devices and storage media. Background [0002] Nowadays, speech recognition technology has been widely used in people's daily lives, bringing great convenience to people's lives. Usually, within the system where speech recognition technology is applied, a speech pronunciation dictionary is pre-established, which includes: the correspondence between each word and the corresponding sequence of pronunciation factors. [0003] However, the phonetic pronunciation dictionary has limited coverage of words, and with the emergence of some emerging words, if the pronunciation of these words is not added to the phonetic pronunciation dictionary in time, the accuracy of speech recognition results will be greatly reduced. Related techniques are to predict the pronunciation of the word by entering a new w...

AI Technical Summary

Problems solved by technology

However, some existing word pronunciation prediction models are obtained based on conventional cyclic neural network training, for example: the hidden layer in the model uses conventional LSTM neurons for abstract learning, when it is necessary to predict the pronunciation of a large number of new words When , the operation speed of the LSTM neuron is limited, and the prediction efficiency of the entire word pronunciation prediction model cannot be guaranteed.

Images