Named entity determination method and device

Abstract

The invention discloses a named entity determination method and device. The method comprises the steps of receiving an input vector representing a to-be-identified named entity; carrying out bidirectional transmission calculation on the input vector by using an LSTM neural network, respectively determining forward information and backward information of the input vector, and generating a first combined vector according to the input vector and the forward information and the backward information thereof; using an RNN neural network to carry out one-way transmission calculation on the first merging vector, and integrating an input vector in the first merging vector and forward information and backward information thereof to generate a second merging vector; performing probability calculationon the second merging vector according to a preset classification label index number; and generating an output result with a classification label index number according to the probability calculationresult. By implementing the method and the device, the relationship between the merged vectors can be determined more accurately, so that in practical application, the recognition accuracy of input vectors such as input texts or natural languages is relatively high.

Description

technical field [0001] The invention relates to the technical fields of machine learning and natural language processing, in particular to a method and device for determining named entities. Background technique [0002] At present, the neural network models involved in the field of machine learning and natural language processing technology mainly include: feedforward neural network models, recurrent neural network models, etc. [0003] The activation function used by the above two neural networks is mainly the sigmoid function, which is used to map the number from negative infinity to positive infinity to between 0 and 1, and the result of deriving this function is f'(x)=f(x )(1-f(x)). Therefore, when multiplying two numbers between 0 and 1, the result will become very small. The backpropagation of the neural network is to multiply the partial derivative of the function layer by layer. Therefore, when the number of layers of the neural network is very deep, the deviation...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the embodiment of the present invention provides a named entity determination method and device to solve the problem of poor recognition accuracy of input vectors such as input text or natural language existing in the prior art

Images