Method for realizing scene reality enhancement in text-to-scene conversion

Abstract

The invention discloses a method for realizing scene reality enhancement in text-to-scene conversion. The method includes the steps of firstly, acquiring multiple Chinese documents describing a scenefrom the Internet to build a scene corpus; secondly, conducting non-duplicate-removal word segmentation on the documents in the corpus, and then conducting stop word removal on the documents after word segmentation; thirdly, conducting statistical analysis on entity nouns in the scene description by using the processed documents; fourthly, analyzing scene category characteristics by using statistical indexes; fifthly, conducting conceptual representation on the scene by using the entity nouns to establish a scene concept dictionary. The method aims at establishing the 'word-category' association between entity words and scene words, analyzing the general characteristics of 'category' and realizing conceptual representation of a certain scene word so as to provide support for analysis of scene entity elements in text-to-scene conversion, thus, the generated scene conforms to the common sense of people and has a complete background environment, and accordingly the scene reality is enhanced.

Description

technical field [0001] The invention belongs to the field of text visualization, and in particular relates to a method for enhancing the authenticity of scenes during text scene conversion. Background technique [0002] Context conversion is a relatively cutting-edge research topic at present. Context conversion basically refers to the ability to convert the symbols of text information into visual analog representations. There are several reasons for the appropriate modeling of this conversion process. A meaningful application is the modeling of human psychological conditions, and another meaningful point is to help understand a story. A third related area is cognitive modeling, applying a large variety of knowledge to interpret a description. Realizing text visualization not only needs to satisfy the description of the text, but also needs to satisfy the actual situation of the scene and the entity. [0003] According to domestic and foreign scientific and technological ...

AI Technical Summary

Problems solved by technology

However, there is no in-depth discussion of the "scene" described in the text

In the existing text-scene conversion system, the input text is mostly simple text, and the generated scene only contains the entities described in the text. The scene does not have obvious category characteristics and background environment, does not have a sense of reality, and the practical significance is not strong

Images