低资源语料下民族文学主题演化预测与追踪算法

打开文本图片集

关键词：低资源语料；民族文学；主题演化预测；迁移学习；元学习；时序注意力机制；特征追踪中图分类号：TN911.2-34；TP18 文献标识码：A 文章编号：1004-373X（2026）06-0120-06

Prediction and trackingalgorithm for evolution of ethnic literature theme under low resource corpus

JING Weiqi'， ZHOU Jingjing²

MinzuUniversityofChina，BeijingOo81，China;2.Xi’anUniversityofPosts&Telecommunications，Xi’an71，Chi

Abstract：Intheapplicationprocessof artificial intellgenceandnaturallanguageprocessing technologytherearemany problemsinthecorpusofnationalliterature，suchasthelackofdigitalresources，thescarcityoflabeledsamplesandsoon. However，thetraditionalthemeanalysismethodshavetheproblemsofinsuficientrepresentationabilityandlowaccuracyof evolutionpredictioninlowresourcescenarios.Therefore，apredictingandtrackingalgorithmfortheevolutionofnational literaturethemeisproposed，whichissuitableforlowresourcecorpus.Thealgorithmintegratesthepreproceingshebased ondomain knowledge，andcanallviate the problemof samplescarcitybymeansofstrategiessuchasdataenhancementand semanticexpansion.Thefusionstrategyofmeta-learningandtransferlearningisused toconstruct thethemerepresentation modelinlowresourcescenarios，whichefectivelysolvestheproblemofinsuficienttextsemanticfeatureextractionundersmall sampleconditions.The proposed topicevolutionpredictionandtrackingalgorithm basedontemporalatention mechanismcan realizeaccuratetrackingof thedynamicevolutionprocessofthethemeandquantitativepredictionoffuturetrends.The experimentalresurtsshowthat，incomparisonwithotheralgorithms，thesubjectrecognitionacuracyandpredictionaccuracyof the proposedalgorithmcanreachabove9O%，whichprovidesafeasibletechnicalpathforthecomputationalanalysisof ethnic literature in low resource scenarios.

Keywords：lowresourcecorpus；ethnic literature;topicevolutionprediction；transfer learning;meta-learning;temporal attention mechanism; feature tracking

0 引言

随着数字人文研究的兴起，民族文学正经历从传统人文解读向计算分析的范式转型。（剩余6497字）