Ancient poem generation method capable of controlling emotion and style

Abstract

The invention provides an ancient poem generation method capable of controlling emotion and style. The method is composed of an ancient poem classification module, an ancient poem generation module and an ancient poem scoring and interaction module. Firstly, ancient poems in a corpus are classified according to styles, emotions and main keyword fine grit, then corresponding labels are determined according to formats, styles, emotions and input texts given by a user, escort poems with controllable styles and emotions are automatically generated, and meanwhile the method can interact with the user to achieve modification and retouching. An encoder pre-training model and a self-attention mechanism with a mask are adopted to ensure automatic association between generated poems and multiple labels. Automatic evaluation and manual evaluation show that the poems with controllable emotions and styles can be generated, and the poem quality exceeds the evaluation baseline level.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, relates to an ancient poem text generation technology, in particular to an ancient poem generation technology with controllable emotion and style. Background technique [0002] The ancient poetry generation system aims to generate user-satisfied rhyming poems according to user needs. First, the poetry writing system must be able to analyze, extract, and summarize keywords suitable for ancient poetry based on the text information provided by users. Generate and correct ancient poems. After the ancient poems are generated, the poetry writing system must also be able to automatically score and interact, and continue to modify and polish the generated ancient poems. [0003] At present, in the field of ancient poetry generation, there are many models that can generate poems that rhyme and meet user needs, but these models can only generate verses based on a single keyword, resulting ...

AI Technical Summary

Problems solved by technology

[0003] At present, in the field of ancient poetry generation, there are many models that can generate rhyming poems that meet user needs, but these models can only generate verses based on a single keyword, resulting in the generated ancient poems.

And these models do not take into account the influence of style and emotion on ancient poems at the same time. Therefore, the verses generated by these models lack diversity, and most of them do not match the user's intentions.

Images