Controllable generative composition embellishing method

The invention discloses a controllable generative composition embellishing method. According to the method, the composition evaluation model of multi-dimensional indexes is utilized, defects of the article are specifically found out, a solution is given, the number of times of directly generating a...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: LUO YANZHUO, ZHANG BO, CHEN XUJIE, MAI YONGQIN, LI XUMIN, LAN WENHUI, DONG ZULAN, ZHANG XIUPING, LIN CHAOCHUN
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention discloses a controllable generative composition embellishing method. According to the method, the composition evaluation model of multi-dimensional indexes is utilized, defects of the article are specifically found out, a solution is given, the number of times of directly generating a complete composition is reduced, and therefore the probability that problems are caused by uncontrollability of natural language generation is reduced. A natural language generation model of fusion planning is adopted, and the content obtained through natural language generation is further controlled. Compared with a traditional computer-aided writing technology, the text modified by the generation type method conforms to language habits, meanwhile, the diversity of natural language expression is not lost, and help is provided for thought development of writers. 本发明公开了一种可控性生成式的作文润色方法。利用了多维度指标的作文评价模型,针对性找出文章的缺陷并给出解决方案,减少了直接生成完整作文的次数,从而降低自然语言生成的不可控性带来问题的概率。采用了融合规划的自然语言生成模型,进一步控制了自然语言生成得到内容。相较于传统计算机辅助写作技术,生成式方法修改后的文本在