Construction Research And System Implementation Of Knowledge Graph Of Cultural Relics In The Context Of Cultural Digitalization

Cultural digitization is to recreate cultural contents through digital technologies such as artificial intelligence algorithms,digital copyright,and audio-visual coding,and to restore,preserve,relate,and disseminate them in digital form.Under the cultural digitization perspective,three types of cultural relics,namely stone carvings,ceramics and bronzes,are selected to build the knowledge graph of cultural relics through domain thesaurus construction,named entity recognition,relationship extraction and other techniques;the knowledge graph system of cultural relics is designed and implemented to reduce the pressure of manual labeling by museum staff through named entity recognition and relationship extraction modules,and to promote the information of cultural relics through visual display modules.The main research contents and innovations of this paper are as follows.(1)The cultural relics dataset and the cultural relics domain lexicon are constructed.The cultural relics dataset includes the cultural relics named entity identification dataset and the cultural relics relationship extraction dataset,which are mainly obtained from the official website of the museum with high reliability.By constructing a seed domain lexicon and expanding the domain lexicon using a new word discovery algorithm,an initial domain lexicon of artifacts is constructed and used as a lexicon in the subsequent named entity recognition part.(2)A lexical enhancement-based recognition algorithm for named entities of artifacts is proposed.The lexical information is introduced in both the input representation layer and the context encoding layer to alleviate the problem of misjudgment of word boundaries in cultural relics data,and has a good experimental effect on the cultural relics dataset.(3)A BERT-based relationship extraction algorithm for cultural relics is proposed.Based on the baseline algorithm of relationship extraction based on BERT,the introduction of entity type information and the optimization of the previous two-entity pairing in the case of multiple entities make up for the shortcomings of the baseline algorithm and achieve better experimental results on cultural relics dataset.(4)A cultural relics knowledge graph system is designed and implemented.The system includes keyword retrieval,named entity recognition algorithm display,relationship extraction algorithm display,and visualization display,which reduces the pressure of manual annotation for museum staff and helps to better explore the connotation of cultural relics and promote the exhibition and dissemination of cultural relics information.In summary,this paper proposes a set of knowledge graph construction methods for cultural relics,and designs and implements a knowledge graph system for cultural relics,which helps to explore the cultural connotation of cultural relics and promote the dissemination of cultural relics information.