Method for converting pictures into Chinese ancient poems based on neural network model

Abstract

The invention discloses a method for converting pictures into Chinese ancient poems based on a neural network model. The method includes the following steps of 1, collecting existing Chinese ancient poems to obtain a poem collection data set; collecting picture resources and sentence description resources corresponding to the picture resources to obtain a picture data set; 2, establishing a multi-mode recurrent neural network and training it to generate picture target description sentences; 3, mapping the target description sentences to Chinese key words; and 4, using long-term and short-termmemory networks to establish a Chinese ancient poem generation model and training the model so as to realize the conversion of the pictures to the Chinese ancient poems. The method automatically converts the pictures into the Chinese ancient poems capable of describing the pictures through a computer, gets rid of the limitation of subject words, and enables an ordinary user to generate a corresponding Chinese ancient poem by inputting a picture, thus, to a certain extent, the method can fill China's vacancy in the field of 'writing poems about pictures' with a machine.

Description

Technical field [0001] The invention relates to the field of information technology, in particular to a method for converting images based on a neural network model to Chinese ancient poems. Background technique [0002] Ancient Chinese poetry is the jewel in the crown of human literature. Since the "Book of Songs" in our country, the poems of the past two thousand years have been shining like stars. Letting machines automatically generate poems has always been a challenging task in the field of artificial intelligence. Humans can easily describe the content of an image, but this task is very difficult for computers. This requires the computer to be able to obtain the content of the image at the semantic level and organize and express the semantic information like humans. [0003] In recent years, deep neural networks have been popular in all directions in the field of artificial intelligence, subverting algorithm design ideas in many fields such as speech recognition, image clas...

AI Technical Summary

Problems solved by technology

[0004] However, the current method of generating ancient poems based on the way of specifying subject words has great limitations. This method has high requirements for the selection of subject words. Only when the subject words are selected reasonably can the generated ancient poems be more reasonable. Many ordinary users cause obstacles; and this method is strictly constrained by the rules and models formulated by some experts, the sentences lack coherence, and the generated target verses are too rigid and lack flexibility

Currently, in the automatic generation of picture descriptions, only simple vernacular sentences are used to capture the content of the pictures for description, and the generated description sentences are monotonous and lack flexibility

Images