Chinese phonetic symbol keyword retrieving method based on feed forward neural network language model

Abstract

The invention provides a Chinese phonetic symbol keyword retrieving method based on a feed forward neural network language model. The method comprises: (1), an input sample including historical words and target words are inputted into a feed forward neural network model; for each target word wi, a plurality of noise words with probability distribution q (wi) are added and an active output of a last hidden layer is transmitted to the target words and nodes where the noise words are located, and conversion matrixes between all layers are calculated based on an objective function; errors between an output of an output layer and the target words are calculated, all conversion matrixes are updated until the feed forward neural network model training is completed; (2), a target word probability of inputting a word history is calculated by using the feed forward neural network model; and (3), the target word probability is applied to a decoder and voice decoding is carried out by using the decoder to obtain word graphs of multiple candidate identification results, the word graphs are converted into a confusion network and an inverted index is generated; and a keyword is retrieved in the inverted index and a targeted key word and occurrence time are returned.

Description

technical field [0001] The invention belongs to the field of speech recognition, and in particular relates to a Chinese speech keyword retrieval method based on a forward neural network language model. Background technique [0002] In the speech keyword retrieval system, the most commonly used language model is the N-gram language model. However, even when N is a small value, such as 3 or 4, the N-gram language model still faces a serious problem of data sparsity. Therefore, many smoothing algorithms are used to alleviate this problem. However, the model’s estimation of the data that does not appear in the training set still has a large gap with the fully trained data, which leads to the inability to correctly identify these uncommon words and the common words around them, which in turn affects the performance of keyword retrieval . The Feed Forward Neural network language model (FFNNLM) maps each word in the dictionary to a continuous space, so it can provide better pred...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome the above-mentioned defects in the current method for reducing the computational complexity of the output layer. By modifying the training criteria of the forward neural network language model, the probability regularization of the output layer is avoided during the training and prediction process, thereby reducing the forward The computational complexity of the neural network language model enables it to be applied to one-pass decoding, thereby improving the retrieval performance of speech keywords

Images