Word order recognition performance analysis method, apparatus and equipment, and storage medium

Abstract

The invention discloses a word order recognition performance analysis method and an apparatus, equipment and a storage medium. The method comprises the steps of obtaining at least one ordered statement and an initial position and an adjusted position of an ordered word in the ordered statement; utilizing a to-be-tested neural network model to convert statement vectors corresponding to the orderedstatements; for each sequenced statement, predicting a first prediction position and a second prediction position of the sequenced word according to the statement vector of the sequenced statement andthe trained position prediction model; and according to the initial positions, the adjusted positions, the first prediction positions and the second prediction positions corresponding to the order adjusting words in the at least one ordered statement, obtaining the position prediction accuracy used for representing the word order information identification performance of the neural network model.According to the scheme, the recognition performance of the neural network on the word order information can be effectively analyzed.

Description

technical field [0001] The present application relates to the technical field of neural networks, and in particular to an analysis method, device, equipment and storage medium for word order recognition performance. Background technique [0002] The neural network model has become a core model in the field of artificial intelligence. For example, neural network models can be applied to natural language processing to implement tasks such as speech recognition or machine translation. [0003] In natural language processing, word order information is a very important information dimension. Among them, word order refers to the sequence of words in a phrase or sentence. It can be seen from this that the ability of the neural network to recognize word order information (commonly known as word order capture ability) is an important dimension for evaluating the processing performance of the neural network for natural language processing. [0004] However, the processing performanc...

AI Technical Summary

Problems solved by technology

[0004] However, the processing performance of neural networks for natural language processing tasks is related to many factors. For example, when using neural networks for machine translation tasks, the translation quality is not only related to word order, but also related to factors such as morphology, semantics, and sentence fluency. , it can be seen that it is difficult to effectively evaluate the word order recognition performance of neural networks based on the processing results of neural networks for natural language processing tasks

Images