Deep Annotation of The Tang Dynasty Seven-Character Quatrain Corpus and Generation of Data Set for Poetry Composition Teaching System
Yulin Yang, Zhifeng Peng · 2023
In the era of AI, the poetry writing system represented by “Jiuge” provides learners with an intelligent platform for teaching poetry creation. Both the previous poetry generation method and the current mainstream deep learning method require large-scale and better quality training corpus as data support. In this study, we take the 304 representative seven-character quatrains of Quan Tang Shi as the corpus source, and complete the following labelling work in accordance with the hierarchy of “lexical meaning (word segmentation, part-of-speech tagging)-imagery-emotion”. Firstly, the deep annotation of words is carried out at the levels of disambiguation, lexical annotation and sense class annotation to complete the description of the semantic hierarchy. Secondly, the labelling of imagery is carried out on the basis of word labelling, showing the relationship between words and imagery through the lexical composition within the imagery. Third, based on the multilayered nature of the semantics of imagery, the literal and non-literal meanings of imagery and the mapping relationship between the two are labelled. Finally, the emotion of the poem is labelled to provide new materials for teaching the appreciation and writing of the quatrains, and to provide refined corpus data for the automatic generation of the quatrains. In this study, 6,122 word examples, 2,615 word types, 5,882 sense class examples, 743 sense types, 427 imagery examples, 296 imagery types, and other in-depth annotated data were completed, which provide a high-quality data base for the AI poetry creation teaching system to carry out training and learning effectively.