Named entity recognition of typical cultural relics focuses on extracting entities from sentences in categories such as name of cultural relic,dynasty,excavation site,and place of collection.The data of typical cultural relics has the specificity of word construction,and using existing named entity recognition methods on typical cultural relics dataset will encounter problems such as wrong word boundary judgments.The algorithm introduces lexical information in both the input representation layer and the contextual encoding layer to improve the word domain expertise.By constructing a lexicon of heritage domain words,the algorithm is used as a lexicon for the lexically enhanced recognition algorithm of typical heritage named entities,which eventually solves the problem of incorrect word boundary judgement and achieves better results on the typical heritage dataset.
维普
万方数据
